Compare commits

...
17 Commits
Author SHA1 Message Date
ketle-manandClaude Opus 5.5 130ba05bd6 chore(release): v0.7.7 - Ollama意思決定モデル対応・Unslothキーなし対応・ヘルプ点検
- docs(help): Video Edit save contents / backup hint, 5 previously
  undocumented Settings sections (helpSettings18-22), font count fix,
  English template fallbacks synced with i18n (EN/JA/ZH)
- docs: DEVLOG
- bump version to 0.7.7

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-30 13:36:43 +09:00
ketle-manandClaude Opus 5.5 fe1b896405 feat(decision): add Ollama 0.35+ as a decision model backend
- decision-client.js: "ollama" backend calls Ollama's /v1/systemone directly
  from the browser (no key, CORS allowed) and surfaces its {error} messages;
  listDecisionModels() lists installed models whose /api/tags capabilities
  include "decision" (tev1, tev1:0.8b, nimble)
- Settings: backend switch (Unsloth / Ollama) updates URL and model list,
  refresh button, pull hint when no decision model is installed; section
  renamed to "Decision Model"
- A model name from the other backend falls back to that backend's default
- docs: help (EN/JA/ZH), README

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-30 13:26:17 +09:00
ketle-manandClaude Opus 5.5 5dbdbd519b feat(unsloth): allow keyless access when UNSLOTH_API_KEY is unset
- Proxy (AI TOOL / sidepanel / Decision API) and the Tagger's Unsloth VLM
  path now send requests without an Authorization header when no key is
  set, for Unsloth Desktop's Keyless API access -> "Chat and inference"
  (localhost-only restriction unchanged)
- A 401 with no key set returns a message explaining both options
- docs: help (EN/JA/ZH), README, .env.example (LAN exposure / keep
  "Allow tools" off)

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-30 13:17:06 +09:00
ketle-manandClaude Opus 5.5 20aff8236b chore(release): v0.7.6 - 意思決定モデル設定・フルバックアップのVideo Edit素材対応
- feat(backup): opt-in "Include media used by Video Edit projects" bundles the
  input-dir clips/BGM referenced by saved projects under _video_edit_media/;
  restored into input/ without overwriting existing files
- fix(backup): stream full-backup uploads to a temp file with read_chunk()
  so restores over ComfyUI's --max-upload-size (default 100 MB) no longer fail
  with "Request Entity Too Large"
- docs: help (Decision Model settings, backup), README, DEVLOG
- bump version to 0.7.6

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-30 12:39:09 +09:00
ketle-manandClaude Opus 5.5 dfe31374cf feat(settings): add dedicated Decision Model (Laya) settings
- New Settings section, stored under wfm_decision_settings independently of
  the AI TOOL / Tagger backends so a decision model can run alongside any
  LLM/VLM: backend (Unsloth for now), URL, model (laya / laya-multilingual /
  laya-english / laya-typed-decisions), default probability threshold
- Test button sends one yes/no question with the unsaved form values and
  shows latency + answer
- decision-client.js reads URL/model/threshold from these settings instead
  of borrowing the AI TOOL backend URL; isYes/pickChoice/pickScore default to
  the saved threshold
- i18n (en/ja/zh)

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-30 12:05:46 +09:00
ketle-manandClaude Opus 5.5 f826b54239 chore(release): v0.7.5 - Ming Image対応・widgets_valuesズレ修正・Unsloth Decision APIクライアント土台
- feat: Unsloth Decision API (/v1/systemone, Laya) relay via the existing
  key-attaching proxy + static/js/decision-client.js (decide/decideMany,
  noul/choice/score builders, probability-threshold helpers)
- docs: README/DEVLOG/help for Ming Image support and the decision client
- bump version to 0.7.5

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-30 11:04:17 +09:00
ketle-manandClaude Opus 5.5 cee7ef1bfa feat: support Ming Image workflows in GenerateUI, Gallery/Metadata and sidepanel I tab
- Detect TextEncodeMingImageEdit prompts (role from sampler/guider wiring,
  propagate role through linked prompt) and EmptyQwenImageLayeredLatentImage
- Settings: disable Sampler field when the SAMPLER source has no sampler_name
  (e.g. SamplerLCM)
- Metadata/I tab: treat BasicGuider "conditioning" as positive; prefer linked
  text source over stale widgets_values[0] in UI-format workflows
- Fix: normalize "legacy full" widgets_values on top-level nodes whose widgets
  are converted to wired inputs (previously only subgraph nodes), which shifted
  later values (e.g. EmptyLatentImage batch_size became 1024)

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-30 10:41:20 +09:00
ketle-manandClaude Opus 5.5 5fb1b1eb01 chore(release): v0.7.4 - Video EditにBGM合成・テキストオーバーレイ・クロップ・Undo/Redoを追加
Video Editサブタブの残りフェーズ(Phase 2〜5)を実装。
- BGM/音声トラック合成: Core音声ノード+ConcatenateVideoのcomplete_audioでグラフ内完結。複数クリップ結合時も各クリップの音声を保持。BGMはボタン/ドロップ/AssetのBGMに設定で指定
- テキストオーバーレイ: クリップ単位の字幕(位置・区間・色・縁取り・背景)、PyAVストリーム処理で焼き込み
- クロップ: Sourceプレビュー上の矩形エディタ(アスペクト固定・キャンセル)、VideoCropで書き出し
- Undo/Redo: 編集状態のスナップショット履歴(ボタン+Ctrl+Z/Y)
- クリアでオーディオ欄も初期化、BGM/Loadのファイル選択が開かない不具合を修正

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-29 15:43:32 +09:00
ketle-manandClaude Opus 5.5 4fa39daf0e docs: document Video Edit BGM, text overlays, crop and undo/redo
README (Video Tab section) and DEVLOG gain an "Unreleased" entry for the
feature/video-edit-bgm-text-overlay branch: audio/BGM track mixing,
per-clip text overlays, crop, undo/redo, BGM from Asset videos / drag &
drop, Clear resetting audio, crop Cancel/auto-exit, and the <label>-based
file picker fix.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-29 15:41:51 +09:00
ketle-manandClaude Opus 5.5 8beec6abdf feat(video-edit): crop edit Cancel button and auto-exit on tab switch
- "Cancel" appears next to "Done" while editing a crop and restores
  the crop the clip had when editing started (including none).
- Switching to the Plan/Asset subtab, another main tab, or another clip
  ends crop editing, keeping the adjusted rect (same as Done). The
  sidebar Project button doesn't, since the Edit subtab stays visible.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-29 12:57:13 +09:00
ketle-manandClaude Opus 5.5 99ffc1fedd feat(video-edit): Clear also resets the Audio section
Clear now empties the clips and restores the soundtrack settings (BGM,
volumes, keep-original, BGM start) to their defaults; it is enabled
whenever either clips exist or audio differs from the defaults, so a
BGM-only timeline can be cleared too. Undo restores both. The confirm
message and help text mention the audio reset.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-29 12:45:04 +09:00
ketle-manandClaude Opus 5.5 a63c6ec53b fix(video-edit): open BGM/Load file pickers via native <label>
The BGM picker (and the project Load button) opened a hidden file input
by calling input.click() from a button's click handler. On the user's
browser this silently showed no dialog, while the Video Source picker in
the same tab — a <label>-wrapped input — worked. Both are now
<label>-wrapped inputs styled as buttons, with Enter/Space wired for
keyboard activation.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-29 12:28:42 +09:00
ketle-manandClaude Opus 5.5 4e39547a4f feat(video-edit): pick BGM from Asset videos and by drag & drop
- Asset subtab: "Set as BGM (audio only)" on video assets uses the
  video's audio track as the Edit BGM (LoadAudio decodes audio from video
  files); videos without an audio track are rejected with a message.
- Audio section accepts a dropped audio/video file, and the BGM file
  input lists common audio extensions and mp4 explicitly instead of only
  audio/*.
- A failed BGM pick no longer clears the previously set BGM.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-29 12:20:46 +09:00
ketle-manandClaude Opus 5.5 c6487ad923 feat(video-edit): add undo/redo for timeline edits
Snapshot history of the editorial state (clip order, trim, crop, text
overlays, audio settings), recorded via a debounced hook on the existing
redraw paths and pushed only when the serialized state changes, so
selection/preview/resize never create entries. Deleted clips come back
from a clip registry without re-uploading. Toolbar buttons plus
Ctrl+Z / Ctrl+Y / Ctrl+Shift+Z (ignored inside text fields); history is
reset on project load.

Also fixes a lost click: committing a trim/hold-length value on blur
rebuilt the text overlay section, replacing a "+ Add text" button that
was being clicked; it now only updates the rows' bounds.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-29 12:04:55 +09:00
ketle-manandClaude Opus 5.5 0c6787f712 feat(video-edit): add per-clip crop with on-preview rectangle editor
Video clips get a normalized crop rect edited directly on the Source
preview (move / corner resize, optional 16:9, 9:16, 1:1, 4:3, 3:4 lock).
Export inserts core VideoCrop (double-nested {"crop": {...}} input, even
pixel values) after Video Slice. The export size is the first video
clip's cropped size; other clips of a different size are center-fit with
ImageScale. Text overlay previews are drawn relative to the cropped
frame, and crops are saved in Edit projects.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-29 12:01:32 +09:00
ketle-manandClaude Opus 5.5 59e96ce180 feat(video-edit): add BGM/audio track mixing and simple text overlays
Phase 4 (audio): the Edit tab export now builds a separate soundtrack in
the same ComfyUI graph and passes it to ConcatenateVideo's complete_audio.
Each clip contributes an exact-length segment (EmptyAudio + AudioMerge with
the clip's own GetVideoComponents audio), chained with AudioConcat, with an
optional BGM (LoadAudio -> TrimAudioDuration -> AudioAdjustVolume) mixed on
top. This also removes the old limitation that multi-clip exports lost
their audio. Export panel gets an Audio section (keep original audio +
volume, BGM file, BGM volume, BGM start offset).

Phase 3 (text): per-clip text overlays (text, start/end, size %, color,
9-grid anchor, outline, background) with a live CSS preview over the
Source/Result panes. Export burns them in via the new
/api/wfm/video/edit/overlay-text endpoint (PyAV streaming decode -> Pillow
composite -> libx264 encode, audio packets remuxed untouched, CJK-capable
system font lookup).

Text overlays and audio settings are persisted in Edit projects.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-29 11:49:55 +09:00
ketle-manandClaude Sonnet 5 6b262a9cf4 chore(release): v0.7.3 - GenerateUIタブにQwen Image 2.1対応を追加
TextEncodeQwenImage21ノード(Comfy-OrgのQwen-Image-2.1テンプレート、T2I/Edit兼用)のprompt/negative_prompt/resolutionをanalyzeWorkflow()で認識し、Input タブのPrompt欄・Settings欄から編集できるようにした。モデルロード・複数参照画像のLoadImageカードは既存の汎用ロジックで無改修対応。

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
2026-09-24 17:23:13 +09:00
25 changed files with 2789 additions and 177 deletions
+5
View File
@@ -3,6 +3,11 @@
# Unsloth Desktop requires an API key even for local access
# (Settings -> API in the Unsloth app, shown only once when issued).
#
# Alternatively leave the key empty and turn on Keyless API access ->
# "Chat and inference" in Unsloth Desktop's Settings -> API; requests are then
# sent without a key. That setting also serves your private LAN, and keep
# "Allow tools" OFF.
#
# 1. Copy this file to ".env" (same directory as this example file)
# 2. Paste your key after the "=" below (do NOT add a leading #, or the
# line becomes a comment and the key will not be loaded)
+5
View File
@@ -2,6 +2,11 @@
Condensed version-by-version index of DEVLOG.md (each line: version + one-line summary of what changed and why). Full rationale, code details, and "How to apply" lessons live in the excluded DEVLOG.md itself; this index exists so graphify can extract a queryable semantic node per release without feeding the full log to the local Ollama model. Regenerate with `python tools/generate_doc_index.py devlog` whenever DEVLOG.md changes.
- **v0.7.7(2026-09-30)** — 意思決定モデル: Ollama 0.35+ バックエンド追加
- **v0.7.6(2026-09-30)** — Settings: 意思決定モデル (Laya) の専用設定
- **v0.7.5(2026-09-30)** — Ming Image ワークフロー対応(GenerateUI / Gallery・Metadata / サイドパネルIタブ)
- **v0.7.4(2026-09-29)** — Video Edit: BGM/音声トラック合成(Phase 4)— 複数クリップ結合時の音声消失も解消
- **v0.7.3(2026-09-24)** — GenerateUIタブにQwen Image 2.1(TextEncodeQwenImage21)対応を追加
- **v0.7.2(2026-09-18)** — Workflowタブにソート機能追加(Date/Name 昇順降順)
- **v0.7.1(2026-09-17)** — セキュリティ修正: Tagger VLM の Unsloth 経路でAPIキーがSSRFにより外部送信され得る不備を修正
- **v0.7.0(2026-09-17)** — プレビュー枠の高さ固定・AssetサブタブをPlan/Editと並ぶ中央下部へ移設・トリムパネルにClipchamp風スクラバー追加・Frame保存のVideo Assets登録とAsset→Plan画像連携
+146
View File
@@ -2,6 +2,152 @@
---
## v0.7.7(2026-09-30)
### 意思決定モデル: Ollama 0.35+ バックエンド追加
Ollama v0.35.0(2026-09-28)が`POST /v1/systemone`で意思決定モデルに対応した(質問・回答の形式はUnsloth/ollayaと同じ)。モデルは`tev1`(Together AI、4B・4.5GB)/ `tev1:0.8b`(812MB)/ `nimble`(Bespoke Labs、9B・9.5GB)で、`ollama pull`で取得する。
- `decision-client.js`: バックエンド`ollama`を追加。APIキー不要でCORSも許可されているため、Unslothのようなサーバー中継は通さずブラウザから直接呼ぶ(既存のOllama LLMバックエンドと同じ)。Ollamaのエラー`{"error": ...}`(例: `model "tev1" not found, try pulling it first`)はそのまま表示。
- `listDecisionModels()`: `/api/tags`の`capabilities`に`"decision"`を持つインストール済みモデルだけを返す(通常のLLMは出ない)。
- Settings「意思決定モデル」: バックエンド切替でURL(11434)とモデル一覧を更新、↻ボタン、意思決定モデルが無い場合は`ollama pull tev1:0.8b`等を案内。Laya専用でなくなったためセクション名を「意思決定モデル」に変更。保存済みモデル名が別バックエンドのもの(Ollamaなのに`laya`等)なら既定モデルへ戻す。
- **実測**: Layaが外した画像プロンプト(「masterpiece, 1girl, silver hair, school uniform…」)の判定を、`tev1:0.8b`は主題person 0.98〜0.99・NSFW「none」0.97〜0.98と全問正解。2回目以降0.2〜0.6秒。ユーザー実機のテストボタンで「接続成功 ✓ 280 ms、yes=0.914」。
- **作業中の不具合**: Pythonスクリプト経由で書いた正規表現の`\b`がバックスペース制御文字(0x08)に化け、モデル名補正が効かなかった(構文エラーにならず単体テストで発覚)。修正後、リポジトリ全体に制御文字が無いことを確認。
### Unsloth: APIキー未設定時はキー無しで送信(Keyless API access対応)
Unsloth Desktopの Settings → API → Keyless API access →「Chat and inference」をオンにするとキー無しで使えるため、`UNSLOTH_API_KEY`未設定を許容した。
- 対象は中継(`py/routes/unsloth_routes.py`: AI TOOL・サイドパネル・意思決定API)とTaggerのUnsloth VLM経路(`py/services/tagger_service.py`)の2か所。キー未設定なら事前エラーにせず`Authorization`無しで送る。localhost限定(SSRF対策)は維持。
- キー無しでUnslothが401を返した場合は「.envにキーを設定するか、Keyless API access →『Chat and inference』をオンに」という案内を返す。
- 偽Unslothサーバー(キー必須/不要)で、キー無し成功・案内メッセージ・キー付与・非localhost拒否を検証(5回連続合格)、ユーザー実機でも確認。
- ヘルプ・README・`.env.example`に、この設定は同じLAN内にもAPIを開放すること、「Allow tools」(キー無しの接続元にPython/ターミナル実行を許す)はオフのままにすることを明記。
- **実機で発生した不具合**: 英語ヘルプ文に`"Chat and inference"`をエスケープせずに入れ、`i18n.js`が構文エラーになりSPA全体が「Loading workflows...」で停止した。この環境では`node --check x.js`がESモジュールの構文エラーを見逃す(終了コード0)ことが原因で、以後は`.mjs`にコピーして`node --check`する方式で確認している。
### ヘルプ(英日中)の点検と修正
Settings・Video(Edit)ページを3言語で点検。言語間の内容差は無かったが、3言語共通の記載漏れ・不正確な点を修正した。
- Video Edit: Saveで保存される内容に、クロップ・テキストオーバーレイ・音声/BGM設定を追記(従来は順序・トリム・参照ファイルのみ記載)。本文に出ていた内部キー名「(see helpVideo9)」を「Projectタブの項目を参照」に。素材はinput内ファイルへの参照のみであること、フルバックアップの素材オプションで保存できることを追記。
- Settings: 未記載だった5セクション(Video Volume、Gallery出力フォルダ、デフォルトCheckpoint、CivitAI APIキー、G'MIC-Qtパス)をhelpSettings18〜22として追加。
- フォント数「Google Fonts 16種」→「15種+システム既定」に訂正。
- テンプレート(index.html)の英語初期表示のうち古くなっていた項目を、Settings・Video全項目についてi18nの英語と同期。
---
## v0.7.6(2026-09-30)
### Settings: 意思決定モデル (Laya) の専用設定
v0.7.5で入れた`decision-client.js`は接続先を「AI TOOLのバックエンドがUnslothならそのURL、それ以外は8888」とAI TOOL設定から借りていたため、LLM/VLMと併用するとURL変更が連動する・モデルやしきい値を選べない問題があった。Settingsタブに独立したセクションを新設し、`wfm_decision_settings`(backend / baseUrl / model / threshold)としてAI TOOL・Tagger設定とは別に保存する。
- backendは現状Unslothのみだが、同じTypeSafe互換APIを話すollaya(:11435、キー不要)等を後から足せるようフィールドとして持つ。
- モデル: `laya`(Unsloth側で選択中)/ `laya-multilingual` / `laya-english` / `laya-typed-decisions`。
- 既定しきい値(0.5〜0.99、既定0.8): `isYes`/`pickChoice`/`pickScore`でしきい値省略時に使う。
- テストボタン: 保存前のフォーム値でnoul1問を送り、応答時間と回答を表示(`testDecisionConnection()`)。
- **実機確認**: 初回はUnsloth側で Settings → API → Decision API → Serve requests がオフのため HTTP 404 `The Decision API is off`(中継・APIキーは正常に通っている)。オン後は「接続成功 ✓ 10430 ms、テスト回答 yes=0.945」(初回はモデル読み込み込み)。
### Settings: フルバックアップのVideo Edit対応
Video Editのプロジェクト(`video_edit_project/*.json`)はデータフォルダ内なので元からZIPに入っていたが、中身はComfyUIの`input/`にアップロードしたクリップ・BGMへの参照(`serverRef`)だけで、素材そのものは含まれていなかった。別環境への復元や`input/`整理後はプロジェクトを開いてもクリップが読み込めない。説明文にもVideo Editの記載がなかった。
- 3つ目のオプション「Video Editプロジェクトの素材を含める」(`include_video_media=1`): 全プロジェクトの`clips[].serverRef`と`audio.bgm.serverRef`(type=input)を`_video_edit_media/<subfolder>/<filename>`に同梱。重複・欠損は除外し、`input/`の外を指す参照(`..`等)は含めない。
- 復元時、`_video_edit_media/`はデータフォルダではなく`input/`へ戻す。同名ファイルが既にあれば上書きしない(`input/`は他のワークフローと共有のため)。結果に`media_restored`/`media_existing`を返し「Video Edit素材: ○件復元、○件は既存」と表示。エクスポート専用の`_external/`とは扱いが異なる。
- 説明文(Settings注記・ヘルプ・README)のバックアップ対象にVideo Editプロジェクトを追加。
**実機で判明した既存問題(Request Entity Too Large)**: aiohttp 3.14の`BodyPartReader.read()`はComfyUIの`--max-upload-size`(既定100MB)を強制する。workflows/wildcardオプション込みのバックアップは約125MB(`_external/`だけで118MB)で超過し復元できなかった(今回の追加以前から起こり得た)。`handle_import_full`を`read_chunk()`で1MBずつ一時ファイルへ流し込む方式に変更し、上限に関係なく復元できるようにした(ZIP全体をメモリに載せることもなくなった)。
**検証**: 一時フォルダでの単体テスト(素材収集・重複/欠損除外・パストラバーサル遮断・既存非上書き)、上限1MBのaiohttpテストサーバー(ComfyUIと同じ3.14.3)への5MBアップロード、ユーザー実機で「インポート完了: 45件 / Video Edit素材: 1件復元、3件は既存」を確認。
### ヘルプ
Settingsページの「Connections」カードに意思決定モデルの項目(helpSettings17、英日中)を追加。フルバックアップの項目(helpSettings15/16)にVideo Editプロジェクトと素材オプションを追記し、テンプレート側の英語初期表示もi18nと同期(15はVideo Plansが抜けたままになっていた)。
---
## v0.7.5(2026-09-30)
### Ming Image ワークフロー対応(GenerateUI / Gallery・Metadata / サイドパネルIタブ)
ユーザー提供の`ming_image_test1〜5.json`(T2I・参照画像Edit・レイヤー分解×2・標準KSampler T2I)に対応。
- `comfyui-workflow.js`: `TextEncodeMingImageEdit`(`prompt`単一ウィジェット+`images.image_N`最大8枚のAutogrow参照画像)を`TextEncodeQwenImageEdit`と同じ扱いで検出。ロールはsampler/guider配線から決まるので、レイヤー分解ワークフローでKSamplerのnegativeに繋がる空プロンプトの2つ目のインスタンスはNegativeとして出る。linked promptへのロール伝播も追加。`EmptyQwenImageLayeredLatentImage`をLatentパネル対象に追加。
- SamplerCustomAdvancedのSAMPLER入力が`sampler_name`を持たないノード(`SamplerLCM`)の場合は`samplerNodeId=null`とし、SettingsパネルのSampler欄をN/A無効化(従来は無関係なsampler名が選択表示され、Applyで存在しない入力に書き込んでいた)。
- `metadata-tab.js` / `node_sets_menu.js`: `BasicGuider`の`conditioning`入力をpositive扱い(従来は判別不能テキスト扱い)。UI形式でCLIPTextEncodeの`text`がリンクされている場合、`widgets_values[0]`に残る古い入力値よりリンク先(`PrimitiveStringMultiline`等)を優先。
**汎用バグ修正(top-levelノードのwidgets_valuesズレ)**: ウィジェットをリンク入力化しても現行フロントエンドは値を`widgets_values`に残す(legacy full形式)。サブグラフ内部ノードは`_stripLegacyLinkedWidgetValues`で正規化していたが、top-levelノードは未対応で、リンクされたウィジェットより後ろの値が1つずつズレていた(`EmptyLatentImage`のwidth/heightを`GetImageSize`/`PrimitiveInt`に繋ぐと`batch_size=1024`、`EmptyQwenImageLayeredLatentImage`の`layers=1024`等)。判定ロジックを`_normalizeLinkedWidgetValues`に切り出し、`convertUiToApi`の本ループ前にも適用。
**検証**: Node上で稼働中ComfyUIの`/object_info`を使い3系統(analyzeWorkflow / metadata-tab / node_sets_menu、UI形式・API形式両方)を5ワークフローで確認。回帰確認として`user/default/workflows`配下366件を修正前後の`convertUiToApi`で比較し、差分35件はすべてズレていた値が正しくなったもの(KSampler `steps=397027830130951`→10、`cfg=25`→8等)。変換エラー1件は修正前から発生する無関係なもの。ユーザー実機で生成まで確認済み。
### Unsloth Decision API(Laya決定モデル)クライアントの土台
Unsloth DesktopがTypeSafe互換の意思決定API(`POST /v1/systemone`、Laya)を提供するようになったため、活用機能の試行に先立って呼び出しの土台のみ実装(UIからの利用は未実装)。
- `py/routes/unsloth_routes.py`: 既存プロキシ(`.env`のAPIキー付与+localhost限定)の`_ALLOWED_PATHS`に`/v1/systemone`を追加。HTTPエラー時はUnsloth側エラー本文の先頭300字をmessageに含める(質問スキーマ誤りをフロントから診断できるように)。
- `static/js/decision-client.js`(新規、SPA用): `decide(state, questions, {model, baseUrl})`・`decideMany`(同時実行数制限、個別失敗は`{error}`)、ビルダー`noul`/`choice`/`score`(Unslothの`instructions`+`criteria`形式)、上限チェック(64問/255選択肢/10段階)、`isYes`/`pickChoice`/`pickScore`。ドキュメントの推奨どおり`confidence`ではなく`probabilities`でしきい値判定し、しきい値未満は`confident:false`(自動適用せず提案表示する用途向け)。接続先はAI TOOL設定がUnslothならそのURL、それ以外は`http://localhost:8888`。
- Layaはテキスト専用(画像入力なし)。画像の判定にはTaggerタグやVLMキャプションを経由させる必要がある。
- 検証はプロキシをモックしたNode単体テストのみ(作業時にUnslothのDecision APIが待ち受けておらず実APIは未検証)。
---
## v0.7.4(2026-09-29)
### Video Edit: BGM/音声トラック合成(Phase 4)— 複数クリップ結合時の音声消失も解消
[VIDEO_EDIT_TAB_PLAN.md](VIDEO_EDIT_TAB_PLAN.md)のPhase 4。着手前に既存の動画編集系ノードを調査し、ComfyUI Core 0.37.0の音声ノード群(`LoadAudio`/`TrimAudioDuration`/`AudioConcat`/`AudioMerge`/`AudioAdjustVolume`/`EmptyAudio`)と`ConcatenateVideo`の`complete_audio`入力(結合動画全体の音声を上書き)だけで、書き出しグラフ内で完結できると判明したため、計画にあった独自`mix-audio` APIは作らなかった。音声トラックの組み方はH3Studioのreel書き出し(クリップ音声を尺にconform→連結→musicを重ねる)を参考にした。
**グラフ構成**: クリップごとに`EmptyAudio(クリップ尺)`+`AudioMerge(自クリップの音声)`で尺ぴったりのセグメントを作り(`AudioMerge`はaudio2をaudio1の長さにpad/trimし、音声なしクリップ=Noneならaudio1をそのまま返す)、`AudioConcat`で連結→(元音量≠0dBなら`AudioAdjustVolume`)→BGM(`LoadAudio→TrimAudioDuration(開始位置, 総尺)→AudioAdjustVolume`)を`AudioMerge`で重ねて`complete_audio`へ渡す。単一の動画クリップで元音声そのまま・BGMなしの場合のみ従来通り再エンコードなしでパススルー。
**UI**: 書き出しパネルにオーディオ欄(元の音声を残す+音量dB、BGMファイル、BGM音量、BGM開始位置)。BGMは「BGM選択...」ボタン、オーディオ欄への音声/動画ファイルのドラッグ&ドロップ、AssetサブタブのVideoに追加した「BGMに設定(音声のみ)」(動画の音声トラックを使う。Galleryは音声ファイルを扱わないため)で指定できる。▶プレビュー中はBGMも`<audio>`で同時再生。クリアボタンでオーディオ欄も初期状態に戻る(BGMのみ設定されている状態でもクリア可能)。
**実機で判明した注意点**: 生の`Video Slice`出力と画像由来の`CreateVideo`クリップを`ConcatenateVideo`へ直接混ぜると`Video chunk N could not be encoded compatibly`(extradata/color space不一致)で失敗する——complete_audioの有無に関係なく。既存の「複数クリップ時は全クリップを`GetVideoComponents→CreateVideo`で再構成する」処理は音声対策だけでなくこのためにも必要だったので維持した。
**検証**: 動画2本(片方トリム)+静止画+BGM(-6dB)をUIから書き出し、尺6.625秒・音声同尺、区間ごとのRMSで元音声+BGM/BGMのみを確認。音声なし動画(audio=None)+BGM、mp4をBGMにした書き出しも確認。
### Video Edit: テキストオーバーレイ(Phase 3)
クリップ単位のテキスト(文字・クリップ内の開始/終了秒・サイズ(フレーム高さ%)・色・9方向アンカー・縁取り・背景)。クリップ相対時刻で保持するので並べ替えに追従し、書き出し時に絶対時刻へ変換する。
Core `TextOverlay`は上/下固定・区間指定なし・既定フォントがCJK非対応、既存のテキスト系カスタムノード(advanced-textoverlay等)も静止画向けか独立パイプラインで、かつ全フレームをIMAGEテンソル化するとメモリが厳しい(720p×720フレームで約8GB)ため、新設の`POST /api/wfm/video/edit/overlay-text`(`VideoService.overlay_text_on_video()`、`asyncio.to_thread`)でPyAVストリーム処理にした: 1フレームずつデコード→該当区間のテキストレイヤー(事前に1回だけ描画)をPillowで合成→libx264でエンコード、音声パケットは再エンコードせずリマックス。書き出しはグラフ出力を`wfm_edit_pre`中間ファイルに保存→焼き込み→中間ファイル削除(`wfm_edit_pre`接頭辞かつoutput内のみ削除対象)の2段構成。フォントはmeiryo→YuGothM→NotoSansJP→msgothic→arialの順に探索。プレビューはSource/Result枠上にCSSで近似表示(`video-preview.js`に`getPreviewPaneElements()`を追加)。
**検証**: ComfyUI再起動後、動画+画像クリップに英日テキスト2件を付けて書き出し、各時刻のフレームで表示/非表示・位置・日本語描画・背景、音声長の一致、中間ファイル削除を確認。
### Video Edit: クロップ(Phase 2)
動画クリップごとに正規化座標(0..1)のクロップ矩形を持ち(ComfyUI-LoadVideoCropと同じ方式)、Sourceプレビュー上で移動+四隅リサイズ(アスペクト比 自由/16:9/9:16/1:1/4:3/3:4固定可、外側は暗転)。「完了」で確定、「キャンセル」で編集前のクロップに戻す。Plan/Assetサブタブ・他タブへの移動、別クリップ選択時は「完了」扱いで編集モードを自動終了。書き出しはCore `VideoCrop`(0.37.0でもexperimental)を`Video Slice`の直後に挿入——`crop`入力は計画どおり二重ネスト`{"crop":{"x","y","width","height"}}`で、偶数ピクセルに丸める(H.264 4:2:0制約)。出力サイズは先頭動画クリップのクロップ後サイズで、サイズの異なる他クリップは`ImageScale(crop=center)`でフィット。解像度不一致ガードはクロップなしの動画クリップ同士だけ比較。`ImageCrop`は0.37でdeprecatedのため不採用。
**検証**: 608x352→300x200のクロップ出力が元フレームの同領域と一致(平均画素差3.1=コーデック誤差程度)。UIで1:1クロップしたクリップを先頭に書き出し、280x280・字幕がクロップ後の枠内・2本目が中央フィットされることを確認。
### Video Edit: Undo/Redo(Phase 5残り)
編集状態(クリップ順・トリム・保持時間・クロップ・テキスト・オーディオ設定)のJSONスナップショット履歴(上限100)。各編集経路に個別に仕込むのではなく、全編集が必ず通る既存の再描画3関数(`_renderTimeline`/`_refreshTextPreview`/`_syncAudioPanel`)から300msデバウンスで記録し、状態が変わったときだけpushする(選択・プレビュー・リサイズでは積まれず、連続入力やドラッグは1ステップにまとまる)。削除クリップはFile参照を持つ`_clipRegistry`から再アップロードなしで復元。↶/↷ボタンとCtrl+Z / Ctrl+Y / Ctrl+Shift+Z(入力欄内はブラウザ標準のundoを優先)。プロジェクト読込時は履歴リセット。
### その他
- **プロジェクト保存**にクロップ・テキスト・オーディオ設定を追加(旧プロジェクトは既定値で読込)。
- **BGM選択・Loadボタンが実ブラウザでダイアログを開かない不具合を修正**: hidden `<input type=file>`をボタンのclickハンドラから`input.click()`で開く方式が、ユーザー環境ではエラーも出さずに無反応だった(同タブのVideo Sourceは`<label>`で包む方式で正常動作)。両ボタンを`<label>`方式に変更。Playwrightはファイル選択を横取りするため自動テストでは再現できなかった。
- **トリム値入力直後に「+ テキスト追加」のクリックが効かない不具合を修正**(テキスト欄実装時に混入): blur時の`change`でテキスト欄全体を再構築し、押そうとしたボタンが置き換わっていた。上限値だけ更新する方式に変更。
**How to apply**: ComfyUI Coreの新しい動画/音声ノードは、`execution_success`で判断せず出力をprobe・画素比較まで実測すること(`VideoTrim`/`VideoCrop`の二重ネストのように、形式を誤ってもエラーにならず無視されるノードがある)。ファイル選択UIは`<label>`で包むネイティブ方式で作り、動作確認はユーザー実機で行うこと。
---
## v0.7.3(2026-09-24)
### GenerateUIタブにQwen Image 2.1(TextEncodeQwenImage21)対応を追加
Comfy-Orgのnightly workflowテンプレート`qwen_image2.1_t2i.json`系(`TextEncodeQwenImage21`ノードでprompt/negative_prompt/resolutionとimage_1〜image_10参照画像を一手に扱う、T2IとEdit兼用のsubgraph構成)をGenerateUIタブで使えるようにした。
**変更内容**: `comfyui-workflow.js`の`analyzeWorkflow()`に`TextEncodeQwenImage21`の分岐を追加。`TextEncodeBooguEdit`と同じ「1ノードにprompt/negative_prompt両方を持ち、ロールは固定(KSampler側の配線をたどらない)」パターンで`prompt_nodes`へ登録し、Input タブのPositive/Negative Prompt欄に反映されるようにした。あわせて`_getWidgetMapping()`に`TextEncodeQwenImage21: ["prompt", "negative_prompt", "resolution"]`を追加し、Settings欄からの書き戻しにも対応。
**調査で判明した既存アーキテクチャ**: GenerateUIのModel/Diffusion Model/Text Encoder/VAE/KSampler/EmptyLatentImage検出は元々ノードタイプ非依存の汎用ロジックなので、UNETLoader+CLIPLoader+VAELoaderの3分割ロード構成(Flux/Qwen Image系で共通)は無改修でそのまま動作した。また画像アップロードUI(Input タブのImageサブパネル)は`generate-tab.js`ではなく`comfyui-editor.js`が担当しており、`analysis.load_image_nodes`に載る全LoadImageノード(1個に限らず)に対して個別カードを描画する汎用実装だったため、Qwen Image 2.1の複数参照画像編集(image_1・image_2をそれぞれ別のLoadImageで接続したワークフロー)もコード変更なしでそのまま動作した——ユーザー提供のI2Iテストワークフローで実機確認済み。
**How to apply**: 新しい「1ノードでprompt+negative_prompt+複数画像入力」型の text-encoder ノードをGenerateUIに対応させる際は、まず`analyzeWorkflow()`のprompt_nodes検出だけ確認すればよい場合が多い——モデルロード・サンプラー・LoadImageカードは大抵ノードタイプに依存しない汎用ロジックが既にカバーしている。画像アップロードUIを探すときは`generate-tab.js`ではなく`comfyui-editor.js`を見ること(両者は別ファイル)。
関連: [[project_v071_unsloth_ssrf_fix_and_outer_gitignore]]
---
## v0.7.2(2026-09-18)
### Workflowタブにソート機能追加(Date/Name 昇順降順)
+6 -6
View File
@@ -26,7 +26,7 @@ A comprehensive workflow, asset management, and generation UI plugin for [ComfyU
- Built-in AI tools (translation and more)
![Workflow Studio](https://img.shields.io/badge/ComfyUI-Custom_Node-blue)
![Version](https://img.shields.io/badge/version-0.7.2-green)
![Version](https://img.shields.io/badge/version-0.7.7-green)
## Features (tab -> feature names)
@@ -37,7 +37,7 @@ Thumbnail / Table views, Thumbnail side panel, Badge filtering, Search, Side pan
One-click capture, Auto-save as thumbnail, Embedded workflow metadata, Auto-import
### GenerateUI Tab (v0.3.5)
6-tab layout, Save button, Input tab, Model tab, Settings tab, Always-visible Raw JSON, Bypass/Mute node handling, One-click generation, Seed control, Style selector, Create Catalog / Catalog buttons, Batch type selector, Batch tab, UI-to-API conversion, Video workflow support, Eagle integration
6-tab layout, Save button, Input tab, Model tab, Settings tab, Always-visible Raw JSON, Bypass/Mute node handling, One-click generation, Seed control, Style selector, Create Catalog / Catalog buttons, Batch type selector, Batch tab, UI-to-API conversion, Video workflow support, Eagle integration, Qwen Image 2.1 support, Ming Image support
### Feeder subtab (v0.3.5 / v0.3.42)
ImageFeeder node control, Image library, Selection management, Presets, Continuous Run loop, Index sync, Seed, WFS_GalleryFeeder node, Node & group selector, Image grid, After Gen modes, Run / Stop controls
@@ -75,11 +75,11 @@ Model Browser, Thumbnail / Table views, Table column sort, Search & Filter, User
### Image Edit Tab (v0.3.65)
Layer-based image editor, Loading images, Open PSD button, Close button, New button, Tools, Layer panel, Duplicate layer (⧉), Layer lock, Text quality, Export, Canvas navigation, Undo / Redo
### Video Tab (v0.5.0 – v0.7.2)
Dedicated video generation UI, independent of GenerateUI, Plan subtab, Base settings row, Edit subtab, Edit project persistence, Asset subtab, Project subtab, Two center previews, First Frame / Last Frame are both optional, Semantic node lookup, not hardcoded IDs, Field highlight color, Load in Video, Frame tab, GIF tab, Playback volume
### Video Tab (v0.5.0 – v0.7.4)
Dedicated video generation UI, independent of GenerateUI, Plan subtab, Base settings row, Edit subtab, Edit: Audio / BGM, Edit: Text overlays, Edit: Crop, Edit: Undo / Redo, Edit project persistence, Asset subtab, Project subtab, Two center previews, First Frame / Last Frame are both optional, Semantic node lookup, not hardcoded IDs, Field highlight color, Load in Video, Frame tab, GIF tab, Playback volume
### AI TOOL Tab (v0.3.14)
4-pane layout, Translation pane, Chat pane, Chat pane — image generation via Tool Calling, Chat pane — generate_image extra parameters, TOOLS pane (VLM), TOOLS pane — shared Chat attachment, TOOLS pane (Wildcards), Chat pane — image-to-image (I2I), Chat pane — SVG generation, no ComfyUI workflow involved, Chat pane — Skills, Settings pane, Unsloth backend, Settings pane — Chat Image Generation, Settings pane — Chat I2I Generation, Settings pane — Generation, Model unload, Settings shared, Backend support, URL security, Lemonade's other endpoints (not integrated)
4-pane layout, Translation pane, Chat pane, Chat pane — image generation via Tool Calling, Chat pane — generate_image extra parameters, TOOLS pane (VLM), TOOLS pane — shared Chat attachment, TOOLS pane (Wildcards), Chat pane — image-to-image (I2I), Chat pane — SVG generation, no ComfyUI workflow involved, Chat pane — Skills, Settings pane, Unsloth backend, Decision API client, Settings pane — Chat Image Generation, Settings pane — Chat I2I Generation, Settings pane — Generation, Model unload, Settings shared, Backend support, URL security, Lemonade's other endpoints (not integrated)
### Workflow Studio Library (ComfyUI Side Panel) (v0.3.9)
Tab layout (W / N / P / M / I / A), W — Workflows tab, N — Nodes tab, M — Models tab, P — Prompts tab, I — Information tab, A — AI TOOL tab, Drag & drop workflows, Drag & drop nodes, Drag & drop prompts, Send to Canvas, Copy prompts, Double-click, Search, ⚙ Theme settings
@@ -157,7 +157,7 @@ Click the **camera icon** (next to the W button) in ComfyUI's top bar to capture
- **[Ollama](https://ollama.com/)** — for AI chat assistant, translation, and VLM features
- **[LM Studio](https://lmstudio.ai/)** — alternative backend for translation and VLM (OpenAI-compatible API)
- **[Lemonade Server](https://lemonade-server.ai/)** — alternative backend for translation, chat, and VLM (OpenAI-compatible API)
- **[Unsloth](https://unsloth.ai/)** — alternative backend for translation, chat, and VLM (OpenAI-compatible API); always requires an API key, set via `UNSLOTH_API_KEY` in a `.env` file — `pip install -r requirements.txt` for `python-dotenv`
- **[Unsloth](https://unsloth.ai/)** — alternative backend for translation, chat, and VLM (OpenAI-compatible API); needs an API key set via `UNSLOTH_API_KEY` in a `.env` file, or Unsloth's Keyless API access → "Chat and inference" turned on — `pip install -r requirements.txt` for `python-dotenv`
- **[Eagle](https://eagle.cool/)** — for auto-saving generated images with metadata
- **[comfyui-mask-editor-one](https://github.com/ketle-man/comfyui-mask-editor-one) (v0.1.9+)** — enables BiRefNet background removal, SAM3 text-prompt segmentation, and ABR stamp-brush library in the Image Edit Mask tool; `birefnet.safetensors` must be placed in `ComfyUI/models/background_removal/` for BiRefNet
- **[psd-tools](https://pypi.org/project/psd-tools/)** (v0.6.0) — enables all PSD (Photoshop) features: Gallery bulk "Create PSD" export, Gallery browsing/thumbnails for `.psd` files, and Image Edit's Save PSD / Open PSD buttons; `pip install -r requirements.txt` (ships prebuilt wheels, no C compiler required); without it, PSD-related actions show an error toast with the install command instead of failing silently
+15 -8
View File
@@ -22,7 +22,7 @@ A comprehensive workflow, asset management, and generation UI plugin for [ComfyU
- Built-in AI tools (translation and more)
![Workflow Studio](https://img.shields.io/badge/ComfyUI-Custom_Node-blue)
![Version](https://img.shields.io/badge/version-0.7.2-green)
![Version](https://img.shields.io/badge/version-0.7.7-green)
## Screenshots
@@ -122,6 +122,8 @@ A comprehensive workflow, asset management, and generation UI plugin for [ComfyU
- **Video workflow support** (v0.5.2) — loading a MiniMax H3 (or other all-in-one video-generation) workflow now works from GenerateUI too, not just the dedicated Video tab: the Input tab's Prompt field picks up its `prompt` widget, the Image tab picks up its LoadImage-based `first_frame` the same way any other I2I workflow does, and generation runs with a longer timeout since video generation takes much longer than a still image; the result preview shows the generated mp4 inline (`<video>`) instead of trying to render it as an image
- **LTX-2.5 / Wan2.2 support** (v0.5.4) — these follow ComfyUI's regular node-graph pattern rather than an all-in-one node, so they needed their own detection: Positive/Negative Prompt now traces through LTX-2.5's `LTXVDualCFGGuider`/`LTXVConditioning` nodes to the actual prompt text, and the Settings tab's Latent Image section recognizes Wan2.2's `EmptyHunyuanLatentVideo` and LTX-2.5's `EmptyLTXVLatentVideo` (width/height/batch size, plus an added **Length (frames)** field) instead of showing "No EmptyLatentImage node found"; the extended generation timeout now also triggers off a `SaveVideo`/`CreateVideo` node anywhere in the graph, not just an all-in-one node
- **Eagle integration** — auto-save generated images to [Eagle](https://eagle.cool/) with metadata
- **Qwen Image 2.1 support** (v0.7.3) — `TextEncodeQwenImage21` (the unified T2I/Edit encoder used by Comfy-Org's Qwen-Image-2.1 workflow templates) is now recognized: its `prompt`/`negative_prompt` widgets populate the Input tab's Positive/Negative Prompt textareas and are editable like any other prompt node; its `resolution` (a pixel-budget int, not width/height) is exposed for Raw JSON editing. No extra work was needed for the model's up to 10 reference images (`image_1`…`image_10`, used for image editing) — each one arrives as an ordinary LoadImage node, so it already gets its own card in the Input tab's Image sub-panel via the existing generic LoadImage detection
- **Ming Image support** (v0.7.5) — `TextEncodeMingImageEdit` (single `prompt` widget, up to 8 optional `images.image_N` reference inputs; role taken from the sampler/guider wiring, so the empty-prompt negative instance in layer-decomposition workflows shows as Negative) and `EmptyQwenImageLayeredLatentImage` (Latent panel; `layers` kept as-is) are recognized in GenerateUI, and Gallery/Metadata/sidepanel I tab now show their prompts as Positive. The Sampler field is disabled (N/A) when the SAMPLER input comes from a node without `sampler_name` (e.g. `SamplerLCM`). **Fix:** top-level nodes whose widgets are converted to wired inputs (e.g. `EmptyLatentImage` width/height fed by `GetImageSize`) keep those values in `widgets_values`; they were not stripped outside subgraphs, shifting later values (`batch_size` became `1024`, KSampler `steps`/`cfg` misread). Also: `BasicGuider`'s `conditioning` is treated as positive, and a linked text input wins over the stale widget text in UI-format metadata
</details>
@@ -236,7 +238,7 @@ An experimental batch generator: runs the workflow currently loaded in GenerateU
- ~~AI Assistant Settings~~ (removed in v0.5.4) — the central Ollama backend/URL/model panel that used to live here was replaced by a ⚙ Settings button + modal directly in the Prompt tab and the Tagger tab, each with its own independent configuration (backend, URL, model, Thinking mode, Max tokens, and a model Unload button) — see those tabs below; the AI TOOL tab's own Settings pane is unaffected
- **Video Playback Volume** (v0.5.4) — the video preview's volume is remembered here and shared across every video player in the plugin (Video tab, Gallery tab), instead of resetting to maximum whenever a different video is opened
- **Data Management** — export core settings/metadata/prompts/CivitAI cache to a single JSON file; import to restore data (useful when migrating or reinstalling); API keys are excluded from exports for security
- **Full Backup (ZIP)** (v0.5.2) — additionally covers what the JSON bundle can't represent: folder-based data (AI Skills, Lab Plans, Video Plans, Style presets) and the Tagger SQLite database; two opt-in checkboxes can also include ComfyUI's default workflows folder (`user/default/workflows`) and the Impact Pack's wildcard folder in the export — both of those are export-only and stored under an `_external/` prefix in the ZIP that Restore always skips, since auto-overwriting files a different plugin/ComfyUI core manages would be unsafe; copy them back by hand if needed. Restoring overwrites existing files; restart ComfyUI afterward for either restore path
- **Full Backup (ZIP)** (v0.5.2) — additionally covers what the JSON bundle can't represent: folder-based data (AI Skills, Lab Plans, Video Plans, Video Edit projects, Style presets) and the Tagger SQLite database; two opt-in checkboxes can also include ComfyUI's default workflows folder (`user/default/workflows`) and the Impact Pack's wildcard folder in the export — both of those are export-only and stored under an `_external/` prefix in the ZIP that Restore always skips, since auto-overwriting files a different plugin/ComfyUI core manages would be unsafe; copy them back by hand if needed. A third opt-in checkbox (v0.7.6) bundles the clips/BGM that saved Video Edit projects reference in ComfyUI's input folder (the project JSON only stores references) under `_video_edit_media/`; unlike `_external/`, these are restored into the input folder, never overwriting a file that already exists there. Restoring overwrites existing files; restart ComfyUI afterward for either restore path
- **Text Size** — one slider (10–28 px) adjusts font size for all prompt and chat textareas at once: Generate UI positive/negative prompts, AI Assistant chat input, Preset prompts, Wildcard prompt and file editor, and Metadata prompt full preview; takes effect immediately and saved with Save Settings
- **RAW JSON Colors** — customize the 6 syntax highlight colors for the Raw JSON editor in Generate UI: Default Text (base), Name/Scheduler (yellow), Title (pink), Width/Height (green), Prompt/Text (cyan), Image/File (red); changes apply immediately on color pick; Reset Defaults restores the original scheme; saved to `localStorage` under `wfm_settings.jsonColors` and applied on startup
- **GenerateUI Model Tab Highlight** (v0.3.98) — pick the color used to mark Checkpoint/VAE/Diffusion Model/Text Encoder/ControlNet/Hypernetwork labels and the LoRA column header in GenerateUI's Model tab when the loaded workflow actually contains that type of node; default bright green; changes apply immediately, Reset Defaults restores it; saved to `localStorage` under `wfm_settings.modelTabActiveColor`; the same color and mechanism also drives the Video tab's field highlighting (v0.5.4, see above)
@@ -414,14 +416,18 @@ An experimental batch generator: runs the workflow currently loaded in GenerateU
</details>
<details>
<summary><h3>Video Tab (v0.5.0 – v0.7.2)</h3></summary>
<summary><h3>Video Tab (v0.5.0 – v0.7.4)</h3></summary>
- **Dedicated video generation UI, independent of GenerateUI** — a purpose-built tab for MiniMax H3 (Image to Video), LTX-2.5 (Text / Image / First & Last Frame to Video), and Wan2.2 (Text to Video) workflows (LTX-2.5/Wan2.2: v0.5.4); does not share state with the GenerateUI tab, and model selection (UNet/CLIP/VAE/Audio VAE) is fixed to the loaded workflow's own defaults rather than editable here
- **Plan subtab** (center panel, batch redesign in v0.5.5) — a **plan** is a sequence of timeline blocks, each a fully self-contained generation job with its own Prompt, Duration, and First/Last Image; split the timeline with **+ Split** / **+ Block** / **Delete**, then **Run** to generate every block in order (Pause/Cancel work mid-run) — a plan with a single block behaves exactly like the old one-shot form. First Image can be None, an explicit image, or "use previous block's last frame" (chaining — an explicit Last Image on the previous block always wins over extracting one from its generated output). Save / Save As / Clear a plan (`ws_videoplan_*.json`, with a PNG contact-sheet thumbnail that embeds the full plan data so dropping the image back on works too)
- **Base settings row** (topbar, v0.5.7) — Aspect Ratio / Megapixels / Multiple / Duration (seconds) / Noise Seed on one row — Duration always reflects the currently selected block (each block keeps its own value), the rest are plan-wide — plus a second row of model-specific fields: Width / Height (become directly editable for workflows with no Resolution Selector node, e.g. Wan2.2 and LTX-2.5's First & Last Frame to Video), Turbo Mode / Strength / Steps (MiniMax H3's turbo LoRA — Turbo Mode doubles as Wan2.2's own turbo toggle, sharing one field), and Prompt Enhance (LTX-2.5 only). Every field in both rows is enabled only when the loaded workflow actually has it — see Field highlight color below (one exception: Wan2.2's Text-to-Video variant doesn't expose Noise Seed at all, so it alone leaves that field disabled)
- **Edit subtab** (v0.7.0) — a lightweight clip timeline for trimming and concatenating video/image clips outside of generation: add clips via the Video Source panel's **Add to Edit** button or the Asset subtab's **Send to Edit** button, drag blocks to reorder (or use the ◀/▶ toolbar buttons), Duplicate/Delete the selected clip, Clear to empty the timeline. The trim panel for a selected video clip shows a Clipchamp-style scrubber — a time ruler, a draggable playhead synced to the Source preview, and draggable in/out handles with time badges (image clips get a simple hold-duration field instead); the playhead's own time badge sits a row above the in/out handle badges (v0.7.2) so it stays legible even when the playhead lands right on a trim point, and the scrubber has extra headroom/height and spacing from the clip name above it (v0.7.2). **▶ Preview** plays every ready clip back-to-back (trimmed to its in/out points) in the result preview, and **Export** builds and runs the actual ComfyUI graph (trim → concatenate if more than one clip → save), tagging the result into Video Temp automatically — exporting more than one clip currently drops audio, since ComfyUI's concatenation node requires uniform audio across all inputs
- **Edit project persistence** (v0.7.2) — **Save** / **Save As** / **Load** buttons in the Edit subtab's export panel persist the timeline itself (clip order, trim points, and which uploaded server-side file each clip refers to) as a `ws_videoeditproj_*.json` project file; Load re-fetches and re-probes each clip from its saved server reference, so the original files need to still exist
- **Asset subtab** (center panel, next to Plan/Edit — moved here from the sidebar in v0.7.0) — browses generated videos/images through three groups: **Video Assets** (a manually-curated keeper group you organize yourself, on your own schedule), **Video Temp** (every plan run's output lands here automatically regardless of whether the plan itself was saved — a **Clear Video Temp** button in the Gallery tab's detail panel empties it in one click), and **All Video Assets** (the union of both); a filename search box, a video/image type filter dropdown (All Types / Video Only / Image Only, v0.7.2), and a thumbnail-grid/table (▦/☰) view toggle sit above the list (v0.7.0). Selecting an item shows it in the center preview and lets you tag/memo it, **Send to Edit** (queues it as a clip in the Edit subtab), jump straight to the full Gallery tab, or — for images only — **Set as First Image** / **Set as Last Image** (v0.7.0, uploads it straight into the currently-selected Plan block and switches to the Plan subtab)
- **Edit subtab** (v0.7.0) — a lightweight clip timeline for trimming and concatenating video/image clips outside of generation: add clips via the Video Source panel's **Add to Edit** button or the Asset subtab's **Send to Edit** button, drag blocks to reorder (or use the ◀/▶ toolbar buttons), Duplicate/Delete the selected clip, Clear to empty the timeline. The trim panel for a selected video clip shows a Clipchamp-style scrubber — a time ruler, a draggable playhead synced to the Source preview, and draggable in/out handles with time badges (image clips get a simple hold-duration field instead); the playhead's own time badge sits a row above the in/out handle badges (v0.7.2) so it stays legible even when the playhead lands right on a trim point, and the scrubber has extra headroom/height and spacing from the clip name above it (v0.7.2). **▶ Preview** plays every ready clip back-to-back (trimmed to its in/out points) in the result preview, and **Export** builds and runs the actual ComfyUI graph (trim → crop → concatenate if more than one clip → save), tagging the result into Video Temp automatically. Multi-clip exports keep each clip's own audio (v0.7.4 — previously they were always silent). **Clear** empties the clips and also resets the Audio section to its defaults (v0.7.4)
- **Edit: Audio / BGM** (v0.7.4) — the export panel's **Audio** section sets the soundtrack: **Keep original audio** (with its own volume in dB) keeps every clip's audio aligned to its segment even across multiple clips (image clips contribute silence), and one **BGM** file is laid over the whole video with its own volume and start offset (a longer BGM is cut to the video length). Pick the BGM with **BGM...**, by dropping an audio/video file onto the Audio section, or with **Set as BGM** on a video in the Asset subtab (that video's audio track is used; videos without audio are rejected). Built entirely inside the export graph with ComfyUI core audio nodes (EmptyAudio / AudioMerge / AudioConcat / LoadAudio / TrimAudioDuration / AudioAdjustVolume) passed to ConcatenateVideo's `complete_audio` — no extra dependencies. ▶ Preview plays the BGM alongside the clips
- **Edit: Text overlays** (v0.7.4) — each clip can carry any number of text overlays (**+ Add text** under the trim panel): text, start/end seconds within the clip, size as % of frame height, color, 9-position anchor, outline, and an optional background box. They follow the clip when it's reordered, are previewed live over the Source/Result panes, and are burned in on Export as a final pass (PyAV decode → Pillow → H.264 encode, frame by frame so memory stays flat; the soundtrack is remuxed untouched). Japanese/CJK text renders with a system CJK font (Meiryo / Yu Gothic / Noto Sans JP on Windows)
- **Edit: Crop** (v0.7.4) — **Edit crop** puts a draggable/corner-resizable rectangle over the Source preview (optionally locked to 16:9 / 9:16 / 1:1 / 4:3 / 3:4); **Done** keeps it, **Cancel** restores the crop from before editing, and switching to another subtab/tab ends editing like Done. Export applies it with ComfyUI core `VideoCrop`; the output size is the first video clip's cropped size, and other clips of a different size are center-fit to it. Text overlay previews are drawn relative to the cropped frame
- **Edit: Undo / Redo** (v0.7.4) — **↶ / ↷** in the timeline toolbar (or Ctrl+Z / Ctrl+Y / Ctrl+Shift+Z outside text fields) step through edits to clips, trims, crops, text and audio settings; a deleted clip comes back without re-uploading. History resets when a project is loaded
- **Edit project persistence** (v0.7.2) — **Save** / **Save As** / **Load** buttons in the Edit subtab's export panel persist the timeline itself (clip order, trim points, and which uploaded server-side file each clip refers to — plus crops, text overlays and audio/BGM settings since v0.7.4) as a `ws_videoeditproj_*.json` project file; Load re-fetches and re-probes each clip from its saved server reference, so the original files need to still exist
- **Asset subtab** (center panel, next to Plan/Edit — moved here from the sidebar in v0.7.0) — browses generated videos/images through three groups: **Video Assets** (a manually-curated keeper group you organize yourself, on your own schedule), **Video Temp** (every plan run's output lands here automatically regardless of whether the plan itself was saved — a **Clear Video Temp** button in the Gallery tab's detail panel empties it in one click), and **All Video Assets** (the union of both); a filename search box, a video/image type filter dropdown (All Types / Video Only / Image Only, v0.7.2), and a thumbnail-grid/table (▦/☰) view toggle sit above the list (v0.7.0). Selecting an item shows it in the center preview and lets you tag/memo it, **Send to Edit** (queues it as a clip in the Edit subtab), **Set as BGM** for videos (v0.7.4 — uses its audio track as the Edit BGM), jump straight to the full Gallery tab, or — for images only — **Set as First Image** / **Set as Last Image** (v0.7.0, uploads it straight into the currently-selected Plan block and switches to the Plan subtab)
- **Project subtab** (sidebar) — lists saved plan files while the center Plan subtab is active, or saved Edit projects while the center Edit subtab is active (v0.7.2 — thumbnail/block-or-clip-count/last-updated) — click one to load it into the matching editor, or delete it with the × button; the panel collapses back down when its own tab is clicked again, and switching Plan↔Edit while it's open refreshes the list in place
- **Two center previews** (v0.5.5) — the left preview shows either the video currently selected in the Asset subtab or whatever's dropped/picked in the **Video Source** panel to its left (**Clear** resets it to empty); the right preview always shows the most recently generated clip; both preview frames are a fixed height (v0.7.0) so switching sources never shifts the Plan/Edit/Asset panel below them. The Frame/GIF tools act on whichever preview was touched most recently
- **First Frame / Last Frame are both optional** — leaving First Image empty runs the workflow text-only (the node is simply left unconnected — a "bypass"); workflows that don't even include a LoadImage node for First Frame are supported out of the box (Wan2.2's Text-to-Video graph has neither slot at all), and a LoadImage node is only injected into the graph on the fly if an image is actually supplied
@@ -450,7 +456,8 @@ An experimental batch generator: runs the workflow currently loaded in GenerateU
- **Chat pane — SVG generation, no ComfyUI workflow involved** (v0.3.77) — the LLM can also produce SVG graphics directly as text; if a reply contains SVG code (wrapped in a markdown code fence or written directly as a bare `<svg>...</svg>`), it's rendered inline as a preview and a "Save to Gallery" button writes it as a `.svg` file to the Gallery tab's output folder; unlike the Tool Calling image generation flow above, this is plain text pattern matching, so it works with any model regardless of tool-calling support
- **Chat pane — Skills** (v0.3.79) — select a `.md` skill file from a dropdown above the message list to prepend its instructions to the conversation as a system prompt (not shown in the visible chat log); a pencil button opens an in-pane manager (list + inline editor, same pattern as the Prompt tab's wildcard file manager) to create, edit, or delete skill files stored in `user/default/Workflow-Studio/ai_skills/`; a file's `---`-delimited frontmatter (`name`/`description`) is shown in the dropdown; two starter skills are bundled — **SVG Icon Generator** (explicitly decomposes the subject into parts before drawing, which markedly improves recognizability from small local models) and **Skill Creator** (interviews you conversationally and drafts a new skill as a ```skill code block; a "Save as new skill" button appears under such replies and opens the manager pre-filled with the draft)
- **Settings pane** — choose backend (Ollama / LM Studio / Lemonade / Unsloth), set the API URL, test connection, select a model (with refresh button), and configure Free language names for translation source and destination
- **Unsloth backend** (v0.3.99) — Unsloth Desktop's OpenAI-compatible API always requires an API key, even for local access; unlike the other backends it's proxied server-side (`py/routes/unsloth_routes.py`, `POST /api/wfm/unsloth/proxy`) so the key never reaches the browser: set `UNSLOTH_API_KEY` in a `.env` file in the plugin folder (copy `.env.example`, loaded via `python-dotenv` in `prestartup_script.py`) and restart ComfyUI; also requires **"Auto model switching (OpenAI API)"** to be turned on in the Unsloth Desktop app's own Settings screen — with it off, requests fail with HTTP 404 unless a model was already manually loaded in the Unsloth Studio UI
- **Unsloth backend** (v0.3.99) — Unsloth Desktop's OpenAI-compatible API requires an API key even for local access unless its **Keyless API access → "Chat and inference"** setting is on; unlike the other backends it's proxied server-side (`py/routes/unsloth_routes.py`, `POST /api/wfm/unsloth/proxy`) so the key never reaches the browser: set `UNSLOTH_API_KEY` in a `.env` file in the plugin folder (copy `.env.example`, loaded via `python-dotenv` in `prestartup_script.py`) and restart ComfyUI — or, since v0.7.7, leave `UNSLOTH_API_KEY` unset and turn on Keyless API access → "Chat and inference" in Unsloth Desktop (requests are then relayed without an `Authorization` header, still only to localhost; that Unsloth setting also opens the API to your private LAN, and "Allow tools" should stay off). A 401 with no key set returns a message explaining both options; also requires **"Auto model switching (OpenAI API)"** to be turned on in the Unsloth Desktop app's own Settings screen — with it off, requests fail with HTTP 404 unless a model was already manually loaded in the Unsloth Studio UI
- **Decision API client** (v0.7.5, foundation) — `static/js/decision-client.js` calls Unsloth Desktop's Decision API (`/v1/systemone`, TypeSafe-compatible Laya decision models) through the same key-attaching server proxy: `decide(state, questions)` / `decideMany`, question builders `noul`/`choice`/`score`, and `isYes`/`pickChoice`/`pickScore` helpers that threshold on `probabilities` (as Unsloth recommends) and flag low-probability answers as not confident. Requires **Decision API → Serve requests** enabled in Unsloth Desktop. Text only — Laya has no image input. **Ollama 0.35+ backend** (v0.7.7) — Ollama's own `/v1/systemone` (decision models `tev1` 4B / `tev1:0.8b` 812 MB / `nimble` 9B, pulled with `ollama pull`; no key, called directly from the browser); the model list shows only installed models whose `capabilities` include `decision`, with a ↻ refresh button and a pull hint when none are installed. Connection settings live in the Settings tab's own **Decision Model** section (v0.7.6): backend (Unsloth / Ollama), URL, model and default probability threshold, stored under `wfm_decision_settings` independently of the AI TOOL / Tagger backends, with a Test button (one yes/no question). No feature uses it yet
- **Settings pane — Chat Image Generation** (v0.3.73) — "Use dedicated workflow" checkbox lets Chat-triggered image generation use a saved workflow of your choice instead of the one currently loaded in GenerateUI, without disturbing what's shown there (same pattern as Image Edit's Inpaint dedicated workflow)
- **Settings pane — Chat I2I Generation** (v0.3.74) — same "Use dedicated workflow" pattern as above, but for image-to-image generation when an image is attached; independent of the text-to-image dedicated workflow setting
- **Settings pane — Generation** (v0.3.98) — **Thinking mode** toggle (off by default; sends Ollama's `think` flag, and regardless of backend, any `<think>...</think>` block is stripped from the output when off so reasoning models don't clutter Translation/Chat/TOOLS results) and **Max tokens** (Ollama: `options.num_predict`; LM Studio/Lemonade/Unsloth: `max_tokens`; `0` = unlimited)
@@ -571,7 +578,7 @@ Click the **camera icon** (next to the W button) in ComfyUI's top bar to capture
- **[Ollama](https://ollama.com/)** — for AI chat assistant, translation, and VLM features
- **[LM Studio](https://lmstudio.ai/)** — alternative backend for translation and VLM (OpenAI-compatible API)
- **[Lemonade Server](https://lemonade-server.ai/)** — alternative backend for translation, chat, and VLM (OpenAI-compatible API)
- **[Unsloth](https://unsloth.ai/)** — alternative backend for translation, chat, and VLM (OpenAI-compatible API); always requires an API key, set via `UNSLOTH_API_KEY` in a `.env` file — `pip install -r requirements.txt` for `python-dotenv`
- **[Unsloth](https://unsloth.ai/)** — alternative backend for translation, chat, and VLM (OpenAI-compatible API); needs an API key set via `UNSLOTH_API_KEY` in a `.env` file, or Unsloth's Keyless API access → "Chat and inference" turned on — `pip install -r requirements.txt` for `python-dotenv`
- **[Eagle](https://eagle.cool/)** — for auto-saving generated images with metadata
- **[comfyui-mask-editor-one](https://github.com/ketle-man/comfyui-mask-editor-one) (v0.1.9+)** — enables BiRefNet background removal, SAM3 text-prompt segmentation, and ABR stamp-brush library in the Image Edit Mask tool; `birefnet.safetensors` must be placed in `ComfyUI/models/background_removal/` for BiRefNet
- **[psd-tools](https://pypi.org/project/psd-tools/)** (v0.6.0) — enables all PSD (Photoshop) features: Gallery bulk "Create PSD" export, Gallery browsing/thumbnails for `.psd` files, and Image Edit's Save PSD / Open PSD buttons; `pip install -r requirements.txt` (ships prebuilt wheels, no C compiler required); without it, PSD-related actions show an error toast with the install command instead of failing silently
+106 -10
View File
@@ -5,13 +5,14 @@ import io
import json
import logging
import shutil
import tempfile
import zipfile
from pathlib import Path
from aiohttp import web
from ..services.settings_service import SettingsService
from ..config import DEFAULT_WORKFLOWS_DIR, DATA_DIR
from ..config import DEFAULT_WORKFLOWS_DIR, DATA_DIR, VIDEO_EDIT_PROJECT_DIR
logger = logging.getLogger(__name__)
@@ -305,7 +306,60 @@ def _add_external_dir_to_zip(zf: zipfile.ZipFile, src_dir: Path, zip_prefix: str
zf.write(path, rel)
def _build_full_backup_zip(include_workflows: bool = False, include_wildcard: bool = False) -> bytes:
# Video Edit projects (video_edit_project/*.json, inside DATA_DIR so always backed up) only
# store *references* — serverRef {filename, subfolder, type:"input"} — to clips and the BGM
# uploaded into ComfyUI's input dir, not the media itself. Backing up just the JSON means a
# restored project on another machine (or after input/ is cleaned up) opens with its clips
# missing. Opt-in (include_video_media) because the media can be large; stored under its own
# prefix so _apply_full_backup_zip can route it back into the input dir, not DATA_DIR.
_VIDEO_MEDIA_PREFIX = "_video_edit_media/"
def _comfyui_input_dir() -> Path:
import folder_paths
return Path(folder_paths.get_input_directory())
def _iter_video_edit_media_refs():
"""Yields (subfolder, filename) for every input-dir file referenced by a saved Video
Edit project (clips + BGM). Unreadable project files are skipped."""
if not VIDEO_EDIT_PROJECT_DIR.is_dir():
return
for path in sorted(VIDEO_EDIT_PROJECT_DIR.glob("*.json")):
try:
with open(path, "r", encoding="utf-8") as f:
data = json.load(f)
except Exception:
continue
refs = [c.get("serverRef") for c in data.get("clips") or [] if isinstance(c, dict)]
bgm = (data.get("audio") or {}).get("bgm")
if isinstance(bgm, dict):
refs.append(bgm.get("serverRef"))
for ref in refs:
if not isinstance(ref, dict) or not ref.get("filename"):
continue
if (ref.get("type") or "input") != "input":
continue
yield (ref.get("subfolder") or "").strip("/\\"), ref["filename"]
def _add_video_edit_media_to_zip(zf: zipfile.ZipFile) -> None:
input_dir = _comfyui_input_dir().resolve()
seen = set()
for subfolder, filename in _iter_video_edit_media_refs():
src = (input_dir / subfolder / filename).resolve()
try:
rel = src.relative_to(input_dir) # a crafted project file must not reach outside input/
except ValueError:
continue
if rel in seen or not src.is_file():
continue
seen.add(rel)
zf.write(src, _VIDEO_MEDIA_PREFIX + str(rel).replace("\\", "/"))
def _build_full_backup_zip(include_workflows: bool = False, include_wildcard: bool = False,
include_video_media: bool = False) -> bytes:
"""Zips everything under DATA_DIR except the excluded names above. Walks
top-level entries first so an excluded symlinked directory (wildcard/) is never
even opened, rather than filtering after a recursive walk has already followed it.
@@ -327,10 +381,12 @@ def _build_full_backup_zip(include_workflows: bool = False, include_wildcard: bo
_add_external_dir_to_zip(zf, _default_workflows_dir(), "_external/default_workflows")
if include_wildcard:
_add_external_dir_to_zip(zf, _impact_pack_wildcard_dir(), "_external/wildcard")
if include_video_media:
_add_video_edit_media_to_zip(zf)
return buf.getvalue()
def _apply_full_backup_zip(zip_bytes: bytes) -> dict:
def _apply_full_backup_zip(zip_source) -> dict:
"""Extracts a full-backup ZIP into DATA_DIR, overwriting existing files. Guards
against Zip Slip (entries whose path resolves outside DATA_DIR via '..' or an
absolute path) by checking each entry before writing it.
@@ -339,17 +395,43 @@ def _apply_full_backup_zip(zip_bytes: bytes) -> dict:
_build_full_backup_zip) are always skipped here rather than restored: they live
outside DATA_DIR and are managed by ComfyUI core / a different custom_node, so
auto-restoring them could silently overwrite files this plugin doesn't own. They
stay in the ZIP for the user to copy back manually if they want them."""
summary = {"extracted": [], "skipped": []}
stay in the ZIP for the user to copy back manually if they want them.
Entries under "_video_edit_media/" (media referenced by Video Edit projects) are restored
into ComfyUI's input dir instead, but never overwrite an existing file there — a file
with the same name is assumed to be the same upload (ComfyUI dedupes uploads by renaming),
and input/ is shared with every other workflow."""
summary = {"extracted": [], "skipped": [], "media_restored": [], "media_existing": []}
DATA_DIR.mkdir(parents=True, exist_ok=True)
base = DATA_DIR.resolve()
with zipfile.ZipFile(io.BytesIO(zip_bytes)) as zf:
input_base = None
# zip_source: raw bytes, or a seekable file object (the upload handler spools to a temp file).
if isinstance(zip_source, (bytes, bytearray)):
zip_source = io.BytesIO(zip_source)
with zipfile.ZipFile(zip_source) as zf:
for info in zf.infolist():
if info.is_dir():
continue
if info.filename.startswith("_external/"):
summary["skipped"].append(info.filename)
continue
if info.filename.startswith(_VIDEO_MEDIA_PREFIX):
if input_base is None:
input_base = _comfyui_input_dir().resolve()
target = (input_base / info.filename[len(_VIDEO_MEDIA_PREFIX):]).resolve()
try:
target.relative_to(input_base)
except ValueError:
summary["skipped"].append(info.filename)
continue
if target.exists():
summary["media_existing"].append(info.filename)
continue
target.parent.mkdir(parents=True, exist_ok=True)
with zf.open(info) as src, open(target, "wb") as dst:
shutil.copyfileobj(src, dst)
summary["media_restored"].append(info.filename)
continue
target = (base / info.filename).resolve()
try:
target.relative_to(base)
@@ -368,11 +450,15 @@ async def handle_export_full(request: web.Request) -> web.Response:
directory, including data the flat JSON bundle (handle_export) can't represent.
Optional query params include_workflows=1 / include_wildcard=1 additionally bundle
ComfyUI's default workflows dir / the Impact Pack's wildcard dir (see
_build_full_backup_zip) — both restore-excluded, export-only, manual-copy-back."""
_build_full_backup_zip) — both restore-excluded, export-only, manual-copy-back.
include_video_media=1 adds the input-dir media referenced by Video Edit projects,
which *is* restored (into input/, without overwriting)."""
try:
include_workflows = request.rel_url.query.get("include_workflows") == "1"
include_wildcard = request.rel_url.query.get("include_wildcard") == "1"
zip_bytes = await asyncio.to_thread(_build_full_backup_zip, include_workflows, include_wildcard)
include_video_media = request.rel_url.query.get("include_video_media") == "1"
zip_bytes = await asyncio.to_thread(
_build_full_backup_zip, include_workflows, include_wildcard, include_video_media)
return web.Response(
body=zip_bytes,
content_type="application/zip",
@@ -390,8 +476,18 @@ async def handle_import_full(request: web.Request) -> web.Response:
field = await reader.next()
if field is None or field.name != "file":
return web.json_response({"error": "file field required"}, status=400)
zip_bytes = await field.read(decode=False)
summary = await asyncio.to_thread(_apply_full_backup_zip, zip_bytes)
# Stream to a temp file with read_chunk() instead of field.read(): aiohttp (3.14+)
# enforces ComfyUI's --max-upload-size (default 100 MB) inside BodyPartReader.read(),
# and a backup with the opt-in workflows/wildcard/Video Edit media easily exceeds it
# ("Request Entity Too Large"). Also avoids holding the whole ZIP in memory.
with tempfile.TemporaryFile() as tmp:
while True:
chunk = await field.read_chunk(1024 * 1024)
if not chunk:
break
await asyncio.to_thread(tmp.write, chunk)
tmp.seek(0)
summary = await asyncio.to_thread(_apply_full_backup_zip, tmp)
return web.json_response({"status": "ok", **summary})
except zipfile.BadZipFile:
return web.json_response({"error": "Not a valid ZIP file"}, status=400)
+27 -11
View File
@@ -5,7 +5,13 @@ for local access (Authorization: Bearer sk-unsloth-...). The key lives in
UNSLOTH_API_KEY (loaded from a .env file by prestartup_script.py) and is
never sent to the frontend — the frontend calls this proxy with the target
path/method/payload, and the server attaches the Authorization header before
relaying to Unsloth's OpenAI-compatible API.
relaying to Unsloth's OpenAI-compatible API — and, via the same relay, to its
Decision API (/v1/systemone, TypeSafe-compatible Laya decision models; see
static/js/decision-client.js).
The key is optional: when UNSLOTH_API_KEY is unset the request is relayed
without an Authorization header, which works when Unsloth Desktop has
Keyless API access → "Chat and inference" turned on (localhost / private LAN).
"""
import json
@@ -20,7 +26,7 @@ from aiohttp import web
logger = logging.getLogger(__name__)
UNSLOTH_DEFAULT_URL = "http://localhost:8888"
_ALLOWED_PATHS = {"/v1/models", "/v1/chat/completions"}
_ALLOWED_PATHS = {"/v1/models", "/v1/chat/completions", "/v1/systemone"}
# baseUrl is client-supplied (so a custom Unsloth port works), but the
# Authorization header carries a real secret — restrict the host to loopback
# so this proxy can't be used to exfiltrate UNSLOTH_API_KEY to an arbitrary
@@ -33,6 +39,14 @@ def _get_api_key():
return os.environ.get("UNSLOTH_API_KEY", "").strip() or None
# Shown when a keyless request is rejected — also used by the Tagger's Unsloth VLM path.
KEYLESS_REJECTED_MESSAGE = (
"Unsloth rejected the request without an API key (HTTP 401). Either set UNSLOTH_API_KEY "
"in the plugin's .env (copy .env.example) and restart ComfyUI, or turn on Keyless API "
"access -> \"Chat and inference\" in Unsloth Desktop's Settings -> API."
)
def _is_allowed_base_url(base_url):
"""Only relay (and attach the API key) to a loopback host."""
try:
@@ -48,9 +62,10 @@ def setup_routes(app: web.Application):
async def handle_proxy(request: web.Request) -> web.Response:
"""POST /api/wfm/unsloth/proxy - Relay a request to Unsloth's OpenAI-compatible API.
"""POST /api/wfm/unsloth/proxy - Relay a request to Unsloth's OpenAI-compatible / Decision API.
Body: { baseUrl, path: "/v1/models" | "/v1/chat/completions", method: "GET" | "POST", payload }
Body: { baseUrl, path: "/v1/models" | "/v1/chat/completions" | "/v1/systemone",
method: "GET" | "POST", payload }
"""
import asyncio
try:
@@ -68,16 +83,13 @@ async def handle_proxy(request: web.Request) -> web.Response:
"message": "Unsloth backend URL must point to localhost/127.0.0.1/::1",
}, status=400)
# No key → relay without Authorization (Unsloth's Keyless API access); if Unsloth
# still demands one, the 401 below explains both ways to fix it.
api_key = _get_api_key()
if not api_key:
return web.json_response({
"message": "UNSLOTH_API_KEY is not set. Copy .env.example to .env in the "
"plugin folder, fill in the key, and restart ComfyUI.",
}, status=401)
def _fetch():
data = json.dumps(payload).encode("utf-8") if payload is not None else None
headers = {"Authorization": f"Bearer {api_key}"}
headers = {"Authorization": f"Bearer {api_key}"} if api_key else {}
if data is not None:
headers["Content-Type"] = "application/json"
req = urllib.request.Request(f"{base_url}{path}", data=data, headers=headers, method=method)
@@ -90,7 +102,11 @@ async def handle_proxy(request: web.Request) -> web.Response:
except HTTPError as e:
detail = e.read().decode("utf-8", errors="replace") if e.fp else str(e)
logger.warning("Unsloth proxy HTTP error: %s %s", e.code, detail)
return web.json_response({"message": f"Unsloth API error: HTTP {e.code}"}, status=e.code)
if e.code == 401 and not _get_api_key():
return web.json_response({"message": KEYLESS_REJECTED_MESSAGE}, status=401)
# Include a short excerpt of Unsloth's own error body — Decision API schema errors
# (e.g. a malformed question) are otherwise impossible to diagnose from the frontend.
return web.json_response({"message": f"Unsloth API error: HTTP {e.code} {detail[:300]}".strip()}, status=e.code)
except URLError as e:
logger.warning("Unsloth proxy connection error: %s", e)
return web.json_response({"message": f"Could not reach Unsloth: {e.reason}"}, status=502)
+31 -2
View File
@@ -1,5 +1,5 @@
"""Video tab API routes: last-frame capture, animated GIF conversion, and
Plan (batch timeline) file management."""
"""Video tab API routes: last-frame capture, animated GIF conversion, Edit
tab probe/text-overlay burn-in, and Plan/Edit project file management."""
import asyncio
import logging
@@ -22,6 +22,7 @@ def setup_routes(app: web.Application):
app.router.add_post("/api/wfm/video/frame/save-to-output", handle_save_frame)
app.router.add_post("/api/wfm/video/to-gif", handle_to_gif)
app.router.add_get("/api/wfm/video/edit/probe", handle_probe_video)
app.router.add_post("/api/wfm/video/edit/overlay-text", handle_overlay_text)
app.router.add_get("/api/wfm/video/plans", handle_list_plans)
app.router.add_get("/api/wfm/video/plans/content", handle_get_plan_content)
@@ -95,6 +96,34 @@ async def handle_probe_video(request: web.Request) -> web.Response:
return web.json_response({"error": str(e)}, status=500)
async def handle_overlay_text(request: web.Request) -> web.Response:
try:
body = await request.json()
filename = body.get("filename", "")
subfolder = body.get("subfolder", "")
type_ = body.get("type", "output")
overlays = body.get("overlays") or []
delete_source = bool(body.get("delete_source"))
if not filename:
return web.json_response({"error": "filename required"}, status=400)
if not isinstance(overlays, list):
return web.json_response({"error": "overlays must be a list"}, status=400)
# Decode+re-encode of every frame — run off the event loop so other
# requests aren't stalled for the duration of the burn-in.
result = await asyncio.to_thread(
_service.overlay_text_on_video, filename, subfolder, type_, overlays, "video/wfm_edit", delete_source
)
return web.json_response({"status": "ok", **result})
except FileNotFoundError as e:
return web.json_response({"error": str(e)}, status=404)
except ValueError as e:
return web.json_response({"error": str(e)}, status=400)
except Exception as e:
logger.error("Error burning text overlay into video: %s", e)
return web.json_response({"error": str(e)}, status=500)
async def handle_list_plans(request: web.Request) -> web.Response:
try:
result = await asyncio.to_thread(_plan_service.list_plans)
+21 -9
View File
@@ -9,6 +9,7 @@ import os
import re
import tempfile
import threading
import urllib.error
import urllib.request
from pathlib import Path
from typing import Optional
@@ -256,9 +257,17 @@ class TaggerService:
# ── VLM Tagging (Ollama / LM Studio / Lemonade / Unsloth) ──
# キー未設定でUnslothに拒否された場合の案内(py/routes/unsloth_routes.pyと同文)
_UNSLOTH_KEYLESS_REJECTED = (
"Unsloth rejected the request without an API key (HTTP 401). Either set UNSLOTH_API_KEY "
"in the plugin's .env (copy .env.example) and restart ComfyUI, or turn on Keyless API "
"access -> \"Chat and inference\" in Unsloth Desktop's Settings -> API."
)
@staticmethod
def _unsloth_api_key() -> Optional[str]:
"""UNSLOTH_API_KEY を環境変数(.env)から取得。Unslothはローカルでも常に必須。"""
"""UNSLOTH_API_KEY を環境変数(.env)から取得。未設定ならNone — その場合はAuthorization無しで
送る(Unsloth DesktopのKeyless API access → "Chat and inference" がオンなら通る)。"""
return os.environ.get("UNSLOTH_API_KEY", "").strip() or None
@staticmethod
@@ -282,15 +291,17 @@ class TaggerService:
logger.error("vlm_models: Unsloth backend URL must point to localhost/127.0.0.1/::1")
return []
key = self._unsloth_api_key()
if not key:
logger.error("vlm_models: UNSLOTH_API_KEY is not set")
return []
headers["Authorization"] = f"Bearer {key}"
if key:
headers["Authorization"] = f"Bearer {key}"
req = urllib.request.Request(api_url.rstrip("/") + "/v1/models", headers=headers)
with urllib.request.urlopen(req, timeout=5) as resp:
data = json.loads(resp.read().decode("utf-8"))
return sorted(m["id"] for m in data.get("data", []))
except Exception as e:
if (backend == "unsloth" and isinstance(e, urllib.error.HTTPError) and e.code == 401
and not self._unsloth_api_key()):
logger.error("vlm_models: %s", self._UNSLOTH_KEYLESS_REJECTED)
return []
logger.error("vlm_models: %s", e)
return []
@@ -324,10 +335,8 @@ class TaggerService:
if not _is_allowed_unsloth_url(api_url):
return {"error": "Unsloth backend URL must point to localhost/127.0.0.1/::1"}
key = self._unsloth_api_key()
if not key:
return {"error": "UNSLOTH_API_KEY is not set. Copy .env.example to .env in the "
"plugin folder, fill in the key, and restart ComfyUI."}
headers["Authorization"] = f"Bearer {key}"
if key:
headers["Authorization"] = f"Bearer {key}"
body = {
"model": model,
"messages": [{"role": "user", "content": [
@@ -360,6 +369,9 @@ class TaggerService:
return {"tags": ", ".join(tags), "count": len(tags)}
except Exception as e:
logger.error("vlm_predict: %s", e)
if (backend == "unsloth" and isinstance(e, urllib.error.HTTPError) and e.code == 401
and not self._unsloth_api_key()):
return {"error": self._UNSLOTH_KEYLESS_REJECTED}
return {"error": str(e)}
# ── ファイルメタデータ書込 ────────────────────────────────
+195
View File
@@ -8,6 +8,8 @@ implementation (e.g. Eagle's video2gif/video-to-frame plugins).
import base64
import logging
import os
import sys
from pathlib import Path
logger = logging.getLogger(__name__)
@@ -18,6 +20,86 @@ try:
except Exception: # pragma: no cover - Pillow is a hard ComfyUI dependency
_RESAMPLE = None
# Text overlay anchors: 9-grid "<vertical>-<horizontal>" (e.g. "bottom-center").
_ANCHOR_V = ("top", "middle", "bottom")
_ANCHOR_H = ("left", "center", "right")
# Fonts tried in order for text overlays. Pillow's built-in default font has no
# CJK glyphs (Japanese/Chinese text would render as boxes), so a system font
# with CJK coverage is preferred; load_default() is only the last resort.
_FONT_CANDIDATES = {
"win32": ["meiryo.ttc", "YuGothM.ttc", "NotoSansJP-Regular.ttf", "msgothic.ttc", "arial.ttf"],
"darwin": [
"/System/Library/Fonts/ヒラギノ角ゴシック W4.ttc",
"/System/Library/Fonts/Hiragino Sans GB.ttc",
"/Library/Fonts/Arial Unicode.ttf",
],
"linux": [
"/usr/share/fonts/opentype/noto/NotoSansCJK-Regular.ttc",
"/usr/share/fonts/noto-cjk/NotoSansCJK-Regular.ttc",
"/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf",
],
}
def _load_overlay_font(size: int):
from PIL import ImageFont
platform = "win32" if sys.platform.startswith("win") else ("darwin" if sys.platform == "darwin" else "linux")
for name in _FONT_CANDIDATES[platform]:
path = name
if platform == "win32" and not os.path.isabs(name):
path = os.path.join(os.environ.get("WINDIR", r"C:\Windows"), "Fonts", name)
if os.path.isfile(path):
try:
return ImageFont.truetype(path, size)
except Exception:
continue
return ImageFont.load_default(size=size)
def _render_text_layer(width: int, height: int, ov: dict):
"""Renders one overlay's text onto a transparent full-frame RGBA layer, so
it can be alpha-composited over every frame in its time range without
re-rendering the text per frame (same idea as ComfyUI core's TextOverlay)."""
from PIL import ImageColor, ImageDraw
text = str(ov.get("text") or "").replace("\\n", "\n")
layer = Image.new("RGBA", (width, height), (0, 0, 0, 0))
if not text.strip():
return layer
size = max(8, round(float(ov.get("font_size") or 6.0) / 100.0 * height))
font = _load_overlay_font(size)
stroke = max(1, round(size * 0.06)) if ov.get("outline", True) else 0
spacing = round(size * 0.2)
try:
color = ImageColor.getrgb(str(ov.get("color") or "#ffffff"))
except ValueError:
color = (255, 255, 255)
fill = (*color[:3], 255)
anchor = str(ov.get("anchor") or "bottom-center").split("-")
v = anchor[0] if anchor and anchor[0] in _ANCHOR_V else "bottom"
h = anchor[1] if len(anchor) > 1 and anchor[1] in _ANCHOR_H else "center"
draw = ImageDraw.Draw(layer)
box = draw.multiline_textbbox((0, 0), text, font=font, spacing=spacing, stroke_width=stroke, align=h)
tw, th = box[2] - box[0], box[3] - box[1]
margin = round(0.04 * min(width, height))
x = {"left": margin, "center": (width - tw) / 2, "right": width - margin - tw}[h] - box[0]
y = {"top": margin, "middle": (height - th) / 2, "bottom": height - margin - th}[v] - box[1]
if ov.get("background"):
pad = round(size * 0.3)
draw.rounded_rectangle(
(x + box[0] - pad, y + box[1] - pad, x + box[2] + pad, y + box[3] + pad),
radius=pad, fill=(0, 0, 0, 140),
)
draw.multiline_text((x, y), text, font=font, fill=fill, spacing=spacing, align=h,
stroke_width=stroke, stroke_fill=(0, 0, 0, 255))
return layer
class VideoService:
def _resolve_media_path(self, filename: str, subfolder: str, type_: str) -> Path:
@@ -157,3 +239,116 @@ class VideoService:
optimize=True,
)
return {"filename": save_name, "subfolder": out_subfolder, "frame_count": len(pil_frames)}
def overlay_text_on_video(
self,
filename: str,
subfolder: str,
type_: str,
overlays: list,
filename_prefix: str = "video/wfm_edit",
delete_source: bool = False,
) -> dict:
"""Burns timed text overlays into a video (Video Edit tab, Phase 3).
ComfyUI core has no time-ranged video text node (core TextOverlay draws
on every frame of an IMAGE batch, top/bottom only), and routing the
whole timeline through IMAGE tensors would hold every decoded frame in
memory at once — so this streams instead: decode one frame, composite
any overlay whose [start, end) covers it, encode, repeat. Audio packets
are remuxed untouched (no audio re-encode), so the soundtrack built by
the export graph (original audio / BGM) survives this pass as-is.
overlays: [{text, start, end, font_size (% of frame height), color,
anchor ("<top|middle|bottom>-<left|center|right>"), outline, background}]
with start/end in seconds from the start of the video.
"""
import av # type: ignore
import folder_paths # type: ignore
src_path = self._resolve_media_path(filename, subfolder, type_)
items = []
for ov in overlays or []:
try:
start = max(0.0, float(ov.get("start") or 0.0))
end = float(ov.get("end") or 0.0)
except (TypeError, ValueError):
continue
if end > start and str(ov.get("text") or "").strip():
items.append((start, end, ov))
if not items:
raise ValueError("No valid text overlays")
output_dir = folder_paths.get_output_directory()
full_output_folder, out_filename, counter, out_subfolder, _ = folder_paths.get_save_image_path(filename_prefix, output_dir)
save_name = f"{out_filename}_{counter:05}_.mp4"
save_path = Path(full_output_folder) / save_name
in_c = av.open(str(src_path))
out_c = None
try:
in_v = in_c.streams.video[0]
in_a = in_c.streams.audio[0] if in_c.streams.audio else None
width, height = in_v.codec_context.width, in_v.codec_context.height
rate = in_v.average_rate or in_v.guessed_rate or 24
layers = [(s, e, _render_text_layer(width, height, ov)) for s, e, ov in items]
out_c = av.open(str(save_path), mode="w")
out_v = out_c.add_stream("libx264", rate=rate)
out_v.width = width
out_v.height = height
out_v.pix_fmt = "yuv420p"
out_v.options = {"crf": "18", "preset": "medium"}
out_a = out_c.add_stream_from_template(in_a) if in_a is not None else None
streams = [in_v] + ([in_a] if in_a is not None else [])
first_pts = None
index = 0
for packet in in_c.demux(*streams):
if packet.stream.type == "audio":
if packet.dts is None:
continue
packet.stream = out_a
out_c.mux(packet)
continue
for frame in packet.decode():
if frame.pts is not None and first_pts is None:
first_pts = frame.pts
t = float((frame.pts - first_pts) * in_v.time_base) if frame.pts is not None else index / float(rate)
active = [layer for s, e, layer in layers if s <= t < e]
if active:
img = frame.to_image().convert("RGBA")
for layer in active:
img.alpha_composite(layer)
out_frame = av.VideoFrame.from_image(img.convert("RGB"))
else:
out_frame = frame
out_frame.pts = index
out_frame.time_base = out_v.codec_context.time_base
index += 1
for out_packet in out_v.encode(out_frame):
out_c.mux(out_packet)
for out_packet in out_v.encode(None):
out_c.mux(out_packet)
except BaseException:
if out_c is not None:
out_c.close()
out_c = None
save_path.unlink(missing_ok=True)
raise
finally:
if out_c is not None:
out_c.close()
in_c.close()
# The export graph's intermediate file is only deleted when the caller
# explicitly asks and it's one of this tab's own "_pre" intermediates in
# output — never an arbitrary user file.
if delete_source and type_ == "output" and Path(filename).name.startswith("wfm_edit_pre"):
try:
src_path.unlink()
except OSError as e:
logger.warning("Could not delete intermediate %s: %s", src_path, e)
return {"filename": save_name, "subfolder": out_subfolder, "frame_count": index}
+1 -1
View File
@@ -1,7 +1,7 @@
[project]
name = "comfyui-workflow-studio"
description = "A comprehensive workflow, asset management, and generation UI plugin for ComfyUI. Manage workflows, models, images, and prompts; batch generation and Image Feeder in the GenerateUI tab; library side panel with canvas drag & drop and AI tools."
version = "0.7.2"
version = "0.7.7"
license = { file = "LICENSE" }
requires-python = ">= 3.10"
dependencies = ["jinja2"]
+232
View File
@@ -602,6 +602,237 @@
flex: none;
}
/* ---- Edit: テキストオーバーレイ(Phase 3) / オーディオ(Phase 4) ---- */
.wfm-video-edit-section-head {
display: flex;
align-items: center;
justify-content: space-between;
gap: 8px;
font-size: 12px;
font-weight: 600;
color: var(--wfm-text-secondary);
margin-bottom: 6px;
}
.wfm-video-edit-text-section {
margin-top: 16px;
padding-top: 10px;
border-top: 1px solid var(--wfm-border);
}
.wfm-video-edit-text-list {
display: flex;
flex-direction: column;
gap: 8px;
}
.wfm-video-edit-text-row {
border: 1px solid var(--wfm-border);
border-radius: 6px;
padding: 6px 8px;
}
.wfm-video-edit-text-row-head {
display: flex;
gap: 6px;
margin-bottom: 6px;
}
.wfm-video-edit-text-row-head input {
flex: 1;
min-width: 0;
}
.wfm-video-edit-text-grid {
display: grid;
grid-template-columns: repeat(auto-fill, minmax(92px, 1fr));
gap: 6px 8px;
align-items: end;
}
.wfm-video-edit-text-grid > label {
display: flex;
flex-direction: column;
gap: 2px;
font-size: 11px;
color: var(--wfm-text-secondary);
margin: 0;
}
.wfm-video-edit-text-grid > label.wfm-video-checkbox-inline {
flex-direction: row;
}
.wfm-video-edit-text-grid input[type="color"] {
width: 100%;
height: 28px;
padding: 0;
border: 1px solid var(--wfm-border);
border-radius: 4px;
background: none;
}
.wfm-video-edit-audio-panel {
margin-top: 10px;
padding-top: 8px;
border-top: 1px solid var(--wfm-border);
}
.wfm-video-edit-audio-row {
display: grid;
grid-template-columns: 1fr auto;
align-items: center;
gap: 2px 6px;
margin-top: 4px;
font-size: 11px;
color: var(--wfm-text-secondary);
}
.wfm-video-edit-audio-row > label {
grid-column: 1 / -1;
margin: 0;
}
.wfm-video-edit-audio-row input[type="range"] {
width: 100%;
}
.wfm-video-edit-audio-row input[type="number"] {
grid-column: 1 / -1;
}
.wfm-video-edit-audio-val {
min-width: 44px;
text-align: right;
font-variant-numeric: tabular-nums;
}
.wfm-video-edit-bgm-file {
display: flex;
align-items: center;
gap: 6px;
margin-top: 6px;
}
.wfm-video-edit-audio-panel.drag-over {
outline: 2px dashed var(--wfm-primary);
outline-offset: 2px;
border-radius: 4px;
}
.wfm-video-edit-bgm-hint {
display: block;
font-size: 10px;
margin-top: 2px;
}
/* <label>で包んだファイル選択を.wfm-btnと同じ見た目にする */
.wfm-video-edit-file-btn {
display: inline-flex;
align-items: center;
justify-content: center;
margin: 0;
user-select: none;
}
.wfm-video-edit-bgm-name {
flex: 1;
min-width: 0;
overflow: hidden;
text-overflow: ellipsis;
white-space: nowrap;
}
.wfm-video-edit-crop-section {
margin-top: 16px;
padding-top: 10px;
border-top: 1px solid var(--wfm-border);
}
.wfm-video-edit-crop-row {
display: flex;
align-items: end;
gap: 12px;
}
.wfm-video-edit-crop-row > label {
display: flex;
flex-direction: column;
gap: 2px;
font-size: 11px;
color: var(--wfm-text-secondary);
margin: 0;
width: 110px;
}
#wfm-video-edit-crop-toggle.active {
background: var(--wfm-primary);
color: #fff;
}
/* Sourceプレビュー上のクロップ矩形。外側はbox-shadowで暗転、
.interactive時のみドラッグ(移動/四隅リサイズ)を受け付ける */
.wfm-video-edit-crop-layer {
position: absolute;
overflow: hidden;
pointer-events: none;
}
.wfm-video-edit-crop-rect {
position: absolute;
box-sizing: border-box;
border: 1px dashed rgba(255, 255, 255, 0.8);
box-shadow: 0 0 0 9999px rgba(0, 0, 0, 0.55);
}
.wfm-video-edit-crop-layer.interactive .wfm-video-edit-crop-rect {
pointer-events: auto;
cursor: move;
border: 2px solid var(--wfm-primary);
touch-action: none;
}
.wfm-video-edit-crop-handle {
display: none;
position: absolute;
width: 12px;
height: 12px;
background: var(--wfm-primary);
border: 1px solid #fff;
border-radius: 2px;
}
.wfm-video-edit-crop-layer.interactive .wfm-video-edit-crop-handle { display: block; }
.wfm-video-edit-crop-handle[data-h="nw"] { left: -7px; top: -7px; cursor: nwse-resize; }
.wfm-video-edit-crop-handle[data-h="ne"] { right: -7px; top: -7px; cursor: nesw-resize; }
.wfm-video-edit-crop-handle[data-h="sw"] { left: -7px; bottom: -7px; cursor: nesw-resize; }
.wfm-video-edit-crop-handle[data-h="se"] { right: -7px; bottom: -7px; cursor: nwse-resize; }
/* プレビュー枠上のテキスト近似表示レイヤー(実際の焼き込みはサーバー側)。
left/top/width/heightはJSが表示中の<video>/<img>の実寸に合わせて設定する */
.wfm-video-edit-text-layer {
position: absolute;
pointer-events: none;
overflow: hidden;
}
.wfm-video-edit-text-item {
position: absolute;
white-space: pre;
line-height: 1.2;
font-family: "Meiryo", "Yu Gothic", "Noto Sans JP", sans-serif;
}
.wfm-video-edit-text-item.outline {
text-shadow: -1px -1px 0 #000, 1px -1px 0 #000, -1px 1px 0 #000, 1px 1px 0 #000, 0 0 2px #000;
}
.wfm-video-edit-text-item.bg {
background: rgba(0, 0, 0, 0.55);
padding: 0.15em 0.3em;
border-radius: 0.3em;
}
.wfm-video-asset-detail label {
font-size: 12px;
color: var(--wfm-text-secondary);
@@ -664,6 +895,7 @@
}
.wfm-video-preview-frame {
position: relative;
flex: 1;
min-width: 0;
height: 260px;
+4
View File
@@ -553,6 +553,10 @@ function applyI18nToHtml() {
"wfm-help-settings-11": "helpSettings11", "wfm-help-settings-12": "helpSettings12",
"wfm-help-settings-13": "helpSettings13", "wfm-help-settings-14": "helpSettings14",
"wfm-help-settings-15": "helpSettings15", "wfm-help-settings-16": "helpSettings16",
"wfm-help-settings-17": "helpSettings17",
"wfm-help-settings-18": "helpSettings18", "wfm-help-settings-19": "helpSettings19",
"wfm-help-settings-20": "helpSettings20", "wfm-help-settings-21": "helpSettings21",
"wfm-help-settings-22": "helpSettings22",
"wfm-help-nodes-1": "helpNodes1", "wfm-help-nodes-2": "helpNodes2",
"wfm-help-nodes-3": "helpNodes3", "wfm-help-nodes-4": "helpNodes4",
"wfm-help-nodes-5": "helpNodes5", "wfm-help-nodes-6": "helpNodes6",
+9 -4
View File
@@ -1382,6 +1382,9 @@ export const comfyEditor = {
// (e.g. Flux2Scheduler has no named "scheduler" combo) so the field is disabled here.
const schedulerAvailable = !!sampler && sampler.schedulerNodeId != null;
const denoiseAvailable = !!sampler && sampler.denoiseNodeId != null;
// Advanced samplers fed by a dedicated sampler node without a sampler_name combo
// (e.g. Ming Image's SamplerLCM) have nothing for the Sampler dropdown to write to.
const samplerNameAvailable = !!sampler && (!sampler.advanced || sampler.samplerNodeId != null);
el.innerHTML = `
<div style="display:flex;flex-direction:row;gap:0;align-items:flex-start;">
@@ -1394,7 +1397,7 @@ export const comfyEditor = {
data-steps-node-id="${sampler.stepsNodeId ?? sampler.id}"
data-cfg-key="${sampler.cfgKey || "cfg"}"
data-cfg-node-id="${sampler.cfgNodeId ?? sampler.id}"
data-sampler-node-id="${sampler.samplerNodeId ?? sampler.id}"
data-sampler-node-id="${samplerNameAvailable ? (sampler.samplerNodeId ?? sampler.id) : ""}"
data-scheduler-node-id="${sampler.schedulerNodeId ?? ""}"
data-denoise-node-id="${sampler.denoiseNodeId ?? ""}">
<div class="wfm-form-group">
@@ -1411,8 +1414,10 @@ export const comfyEditor = {
</div>
<div class="wfm-form-group">
<label>Sampler</label>
<select class="wfm-select" id="wfm-settings-sampler-name">
${this.models.samplers.map((s) => `<option value="${s}" ${s === sampler.sampler_name ? "selected" : ""}>${s}</option>`).join("")}
<select class="wfm-select" id="wfm-settings-sampler-name" ${samplerNameAvailable ? "" : "disabled"}>
${samplerNameAvailable
? this.models.samplers.map((s) => `<option value="${s}" ${s === sampler.sampler_name ? "selected" : ""}>${s}</option>`).join("")
: `<option value="">N/A</option>`}
</select>
</div>
<div class="wfm-form-group">
@@ -1490,7 +1495,7 @@ export const comfyEditor = {
if (stepsEl?.value !== "") write(ds.stepsNodeId, "steps", parseInt(stepsEl.value) || 20);
const cfgEl = document.getElementById("wfm-settings-cfg");
if (cfgEl?.value !== "") write(ds.cfgNodeId, ds.cfgKey || "cfg", parseFloat(cfgEl.value) || 7);
write(ds.samplerNodeId, "sampler_name", document.getElementById("wfm-settings-sampler-name")?.value);
if (ds.samplerNodeId) write(ds.samplerNodeId, "sampler_name", document.getElementById("wfm-settings-sampler-name")?.value);
if (ds.schedulerNodeId) write(ds.schedulerNodeId, "scheduler", document.getElementById("wfm-settings-scheduler")?.value);
const denoiseEl = document.getElementById("wfm-settings-denoise");
if (ds.denoiseNodeId && denoiseEl?.value !== "") write(ds.denoiseNodeId, "denoise", parseFloat(denoiseEl.value) || 1.0);
+64 -12
View File
@@ -291,15 +291,26 @@ function _stripLegacyLinkedWidgetValues(objectInfo, remappedNode, sgDef, origInt
const allWidgetInputs = (remappedNode.inputs || []).filter((i) => i.widget);
if (!widgetsValues || allWidgetInputs.length === 0) return;
const widgetNames = _getWidgetInputNames(objectInfo, remappedNode.type);
const widgetTypes = _getWidgetInputTypes(objectInfo, remappedNode.type);
const isLinkedName = (name) => _isLinkedWidgetName(remappedNode, sgDef, origInternalId, redirectedTargets, name);
const comboExpander = (name, selectedKey) => _getDynamicComboSubNames(objectInfo, remappedNode.type, name, selectedKey);
remappedNode.widgets_values = _normalizeLinkedWidgetValues(objectInfo, remappedNode.type, widgetsValues, isLinkedName);
}
/**
* Return widgets_values with the entries belonging to linked widgets removed when the array is in
* the "legacy full" format (every widget present, linked ones included — what the current ComfyUI
* frontend actually saves for a widget converted to a wired input, e.g. EmptyLatentImage
* [1024, 2048, 1] with width/height fed by GetImageSize/PrimitiveInt). Returns the array unchanged
* when it's already in the modern format (linked widgets contribute zero entries).
*/
function _normalizeLinkedWidgetValues(objectInfo, nodeType, widgetsValues, isLinkedName) {
const widgetNames = _getWidgetInputNames(objectInfo, nodeType);
const widgetTypes = _getWidgetInputTypes(objectInfo, nodeType);
const comboExpander = (name, selectedKey) => _getDynamicComboSubNames(objectInfo, nodeType, name, selectedKey);
// If skipping linked widgets already accounts for every entry, widgets_values is already in
// the modern (no dangling linked entries) format — nothing to strip.
const skipSim = _simulateWidgetValues(widgetNames, widgetTypes, widgetsValues, isLinkedName, true, comboExpander);
if (skipSim.consumed === widgetsValues.length) return;
if (skipSim.consumed === widgetsValues.length) return widgetsValues;
// Otherwise it's "legacy full": recompute with every widget present to find exactly which
// indices the linked ones occupy — including any control_after_generate extra a linked
@@ -311,9 +322,7 @@ function _stripLegacyLinkedWidgetValues(objectInfo, remappedNode, sgDef, origInt
for (const i of fullSim.consumedIndices[name] || []) linkedIdx.add(i);
}
}
if (linkedIdx.size > 0) {
remappedNode.widgets_values = widgetsValues.filter((_, i) => !linkedIdx.has(i));
}
return linkedIdx.size > 0 ? widgetsValues.filter((_, i) => !linkedIdx.has(i)) : widgetsValues;
}
/**
@@ -692,7 +701,14 @@ export const comfyWorkflow = {
});
// 2. Map widgets_values to widget input names using object_info
const widgets = node.widgets_values || [];
let widgets = node.widgets_values || [];
// Widgets converted to wired inputs (e.g. EmptyLatentImage width/height fed by
// GetImageSize) usually still keep their value in widgets_values — strip those so the
// loop below, which assumes linked widgets contribute zero entries, doesn't shift every
// later widget (otherwise batch_size silently becomes the stale width value).
if (Array.isArray(widgets) && widgets.length > 0 && linkedSlotNames.size > 0 && !node.properties?.__lm_widget_ids) {
widgets = _normalizeLinkedWidgetValues(objectInfo, node.type, widgets, (name) => linkedSlotNames.has(name));
}
if (widgets.length > 0) {
// Lora Loader (LoraManager): use __lm_widget_ids from node properties
// widgets_values: [autocomplete_meta_obj, text_str, loras_array]
@@ -1085,7 +1101,7 @@ export const comfyWorkflow = {
// TextEncodeQwenImageEditPlus/TextEncodeQwenImageEdit — propagate role through its
// linked "prompt" input when prompt-enhancement wiring feeds it via a switch chain
// instead of a literal string (e.g. Krea-2's image-reference workflow).
if (ct === "TextEncodeQwenImageEditPlus" || ct === "TextEncodeQwenImageEdit") {
if (ct === "TextEncodeQwenImageEditPlus" || ct === "TextEncodeQwenImageEdit" || ct === "TextEncodeMingImageEdit") {
const tv = inputs.prompt;
if (Array.isArray(tv)) {
if (isPos) samplerPositiveRef[tv[0]] = true;
@@ -1167,7 +1183,11 @@ export const comfyWorkflow = {
cfg: typeof guiderInputs[guiderCfgKey] === "number" ? guiderInputs[guiderCfgKey] : undefined,
cfgNodeId: guiderCfgKey in guiderInputs ? guiderId : null,
cfgKey: guiderCfgKey,
sampler_name: samplerSelInputs.sampler_name, samplerNodeId: samplerSelId,
// samplerNodeId is null when the SAMPLER source has no sampler_name combo at all
// (e.g. Ming Image's SamplerLCM, a dedicated sampler node with its own
// s_noise knobs instead of KSamplerSelect) — the Settings UI disables the field.
sampler_name: samplerSelInputs.sampler_name,
samplerNodeId: "sampler_name" in samplerSelInputs ? samplerSelId : null,
scheduler: undefined, schedulerNodeId: "scheduler" in schedulerInputs ? schedulerId : null,
denoise: typeof schedulerInputs.denoise === "number" ? schedulerInputs.denoise : undefined,
denoiseNodeId: "denoise" in schedulerInputs ? schedulerId : null,
@@ -1270,7 +1290,11 @@ export const comfyWorkflow = {
// the base variant takes one — both share the same "prompt" widget). Only add when
// "prompt" is a direct string; if linked (e.g. Krea-2's prompt-enhancement wiring),
// the upstream literal node is resolved instead via the role propagation above.
if (ct === "TextEncodeQwenImageEditPlus" || ct === "TextEncodeQwenImageEdit") {
// TextEncodeMingImageEdit (Ming Image) has the same single "prompt" widget plus up to
// 8 optional images.image_N reference inputs (none wired = pure T2I); its role comes
// from the sampler/guider wiring — Ming's layer-decomposition workflows use a second,
// empty-prompt instance of it as the KSampler negative.
if (ct === "TextEncodeQwenImageEditPlus" || ct === "TextEncodeQwenImageEdit" || ct === "TextEncodeMingImageEdit") {
const textVal = inputs.prompt;
if (typeof textVal === "string") {
result.prompt_nodes.push({
@@ -1296,6 +1320,28 @@ export const comfyWorkflow = {
}
}
// TextEncodeQwenImage21 (Qwen Image 2.1 unified T2I/Edit encoder) — prompt +
// negative_prompt in one node (fixed roles, same shape as TextEncodeBooguEdit).
// Also carries a "resolution" pixel-budget int (not width/height) and up to 10
// optional images.image_N reference inputs (image_1 is the edit target when wired;
// with no images connected the node behaves as pure T2I). The reference images
// themselves need no special handling here — they arrive as ordinary LoadImage
// nodes already picked up generically by the load_image_nodes pass below.
if (ct === "TextEncodeQwenImage21") {
if (typeof inputs.prompt === "string") {
result.prompt_nodes.push({
id, type: ct, title: `${title} [positive]`, role: "positive",
text: inputs.prompt, textKey: "prompt",
});
}
if (typeof inputs.negative_prompt === "string") {
result.prompt_nodes.push({
id, type: ct, title: `${title} [negative]`, role: "negative",
text: inputs.negative_prompt, textKey: "negative_prompt",
});
}
}
// Mage-Flow text encoder — prompt + negative_prompt in one node (fixed roles,
// unlike KSampler-derived getRole() since both wires originate from this node)
if (ct === "TextEncodeMageFlowEdit") {
@@ -1383,7 +1429,10 @@ export const comfyWorkflow = {
// GetImageSize node (auto-follows the reference image) rather than a direct number —
// leave those as undefined, same treatment as TextEncodeMageFlowEdit below, so the
// settings-tab UI shows "linked" instead of a stale default and Apply doesn't clobber it.
if (ct === "EmptyLatentImage" || ct === "EmptySD3LatentImage" || ct === "EmptyFlux2LatentImage") {
// EmptyQwenImageLayeredLatentImage (Qwen-Image-Layered / Ming Image layer decomposition)
// has the same width/height/batch_size shape plus a "layers" count, left as-is.
if (ct === "EmptyLatentImage" || ct === "EmptySD3LatentImage" || ct === "EmptyFlux2LatentImage"
|| ct === "EmptyQwenImageLayeredLatentImage") {
result.latent_nodes.push({
id, type: ct, title,
width: typeof inputs.width === "number" ? inputs.width : undefined,
@@ -1651,6 +1700,9 @@ function _getWidgetMapping(nodeType) {
ImpactWildcardProcessor: ["wildcard_text"],
ImpactWildcardEncode: ["wildcard_text"],
WFS_PromptText: ["positive", "negative"],
TextEncodeQwenImage21: ["prompt", "negative_prompt", "resolution"],
TextEncodeMingImageEdit: ["prompt"],
EmptyQwenImageLayeredLatentImage: ["width", "height", "layers", "batch_size"],
};
return mappings[nodeType] || null;
}
+252
View File
@@ -0,0 +1,252 @@
/**
* Decision model client (Unsloth Decision API / Laya, Ollama 0.35+ / tev1・nimble).
*
* Laya is a *decision* model, not a generator: it takes a state (text or any JSON) plus typed
* questions and returns calibrated probabilities in one forward pass — no free-form output, so no
* format hallucination, and results can be thresholded directly. Text only (no image input):
* image-based judgments must first be turned into text (Tagger tags / VLM caption).
*
* Both backends speak the same TypeSafe-compatible POST /v1/systemone. Unsloth needs an API key
* even on localhost (unless Keyless API access is on), so its requests go through the same
* server-side proxy as the Unsloth chat backend (py/routes/unsloth_routes.py, key from .env).
* Ollama needs no key and allows CORS, so it's called directly from the browser like the existing
* Ollama LLM backend. Backend/URL/model/threshold come from the Settings tab's own
* "Decision Model" section (wfm_decision_settings), independent of the AI TOOL / Tagger backends so
* a decision model can run alongside whichever LLM/VLM those use.
* API reference: https://unsloth.ai/docs/models/decision-laya,
* https://github.com/ollama/ollama/releases/tag/v0.35.0
*
* Usage (threshold defaults to the Settings tab value when omitted):
* import { decide, noul, choice, score, isYes, pickChoice } from "./decision-client.js";
* const answers = await decide(promptText, {
* genre: choice("Which genre fits best?", { portrait: "a person is the main subject", landscape: "scenery" }),
* text: noul("Does the image contain rendered text?"),
* nsfw: score("How sexually explicit is this?", ["none", "suggestive", "explicit"]),
* });
* const genre = pickChoice(answers.genre, 0.8); // { value, probability, confident }
*/
import { unslothProxy, readJsonStorage } from "./util.js";
// Limits documented by Unsloth's Decision API.
export const DECISION_LIMITS = { maxQuestions: 64, maxChoiceOptions: 255, maxScoreLevels: 10 };
// Unsloth: "laya" = the model picked in Unsloth's settings. Explicit variants: "laya-multilingual"
// (default download, 100+ languages), "laya-english", "laya-typed-decisions" (structured data such
// as JSON records — e.g. workflow node summaries).
// Ollama: decision models are ordinary pulled models whose /api/tags capabilities include
// "decision" — listed live by listDecisionModels(); `models` below are only pull suggestions
// (tev1 = 4B / 4.5 GB, tev1:0.8b = 812 MB, nimble = 9B / 9.5 GB).
export const DECISION_BACKENDS = {
unsloth: {
label: "Unsloth",
defaultUrl: "http://localhost:8888",
models: ["laya", "laya-multilingual", "laya-english", "laya-typed-decisions"],
defaultModel: "laya",
},
ollama: {
label: "Ollama",
defaultUrl: "http://localhost:11434",
models: ["tev1", "tev1:0.8b", "nimble"],
defaultModel: "tev1",
},
};
export const DECISION_MODELS = DECISION_BACKENDS.unsloth.models;
export const DEFAULT_DECISION_MODEL = DECISION_BACKENDS.unsloth.defaultModel;
export const DECISION_SETTINGS_KEY = "wfm_decision_settings";
const DEFAULT_SETTINGS = {
backend: "unsloth",
baseUrl: DECISION_BACKENDS.unsloth.defaultUrl,
model: DEFAULT_DECISION_MODEL,
threshold: 0.8,
};
/** Saved Decision Model settings merged over the defaults. */
export function getDecisionSettings() {
const saved = readJsonStorage(DECISION_SETTINGS_KEY);
const merged = { ...DEFAULT_SETTINGS, ...saved };
const th = Number(merged.threshold);
merged.threshold = th > 0 && th <= 1 ? th : DEFAULT_SETTINGS.threshold;
if (!DECISION_BACKENDS[merged.backend]) merged.backend = DEFAULT_SETTINGS.backend;
const backend = DECISION_BACKENDS[merged.backend];
if (!merged.baseUrl) merged.baseUrl = backend.defaultUrl;
// A model name from the other backend (e.g. "laya" after switching to Ollama) can't work.
const isLaya = /^laya/.test(merged.model || "");
if (!merged.model || (merged.backend === "ollama" && isLaya) || (merged.backend === "unsloth" && !isLaya)) {
merged.model = backend.defaultModel;
}
return merged;
}
/**
* Model choices for the Settings dropdown. Unsloth: its fixed Laya names. Ollama: installed models
* whose /api/tags capabilities include "decision" (throws if Ollama can't be reached). Returns
* { models, installed } — installed is false for Unsloth (names aren't checked against the server).
*/
export async function listDecisionModels(backend, baseUrl) {
if (backend !== "ollama") return { models: [...DECISION_BACKENDS.unsloth.models], installed: false };
const res = await fetch(`${(baseUrl || DECISION_BACKENDS.ollama.defaultUrl).replace(/\/+$/, "")}/api/tags`);
if (!res.ok) throw new Error(`HTTP ${res.status}`);
const data = await res.json();
const models = (data.models || [])
.filter((m) => Array.isArray(m.capabilities) && m.capabilities.includes("decision"))
.map((m) => m.name)
.sort();
return { models, installed: true };
}
export function saveDecisionSettings(patch) {
const data = { ...getDecisionSettings(), ...patch };
localStorage.setItem(DECISION_SETTINGS_KEY, JSON.stringify(data));
return data;
}
// ---- Question builders ----
/** Yes/no question — answer.noul is the probability of "yes" (0..1). */
export function noul(instructions) {
return { type: "noul", instructions };
}
/**
* Pick-one question. criteria: an array of option names, or { option: description } — descriptions
* noticeably improve accuracy (options share a token budget, so past ~20 described options the
* descriptions get trimmed).
*/
export function choice(instructions, criteria) {
return { type: "choice", instructions, criteria };
}
/** Rating question. levels: ordered labels from lowest (0) to highest, 2..10 entries. */
export function score(instructions, levels) {
return { type: "score", instructions, criteria: levels };
}
// ---- Request ----
function _validateQuestions(questions) {
const entries = Object.entries(questions || {});
if (entries.length === 0) throw new Error("decide(): no questions given");
if (entries.length > DECISION_LIMITS.maxQuestions) {
throw new Error(`decide(): ${entries.length} questions exceeds the limit of ${DECISION_LIMITS.maxQuestions}`);
}
for (const [key, q] of entries) {
if (!q || !["noul", "choice", "score"].includes(q.type)) {
throw new Error(`decide(): question "${key}" has an unknown type`);
}
const n = Array.isArray(q.criteria) ? q.criteria.length : Object.keys(q.criteria || {}).length;
if (q.type === "choice" && (n < 2 || n > DECISION_LIMITS.maxChoiceOptions)) {
throw new Error(`decide(): choice "${key}" needs 2-${DECISION_LIMITS.maxChoiceOptions} options (got ${n})`);
}
if (q.type === "score" && (n < 2 || n > DECISION_LIMITS.maxScoreLevels)) {
throw new Error(`decide(): score "${key}" needs 2-${DECISION_LIMITS.maxScoreLevels} levels (got ${n})`);
}
}
}
/**
* Ask typed questions about one state. Returns the raw `answers` object keyed like `questions`.
* The first call after Unsloth starts loads the model (10-20 s); later calls are sub-second on CPU.
*
* @param {string|object} state Text or any JSON (e.g. { prompt, tags, model }).
* @param {object} questions { key: noul(...) | choice(...) | score(...) }, max 64.
* @param {object} [opts] { backend, model, baseUrl } — override the saved Decision Model settings.
*/
export async function decide(state, questions, opts = {}) {
_validateQuestions(questions);
const settings = getDecisionSettings();
const backend = opts.backend || settings.backend;
const baseUrl = (opts.baseUrl || settings.baseUrl).replace(/\/+$/, "");
const payload = { model: opts.model || settings.model, state, questions };
let data;
if (backend === "ollama") {
const res = await fetch(`${baseUrl}/v1/systemone`, {
method: "POST",
headers: { "Content-Type": "application/json" },
body: JSON.stringify(payload),
});
data = await res.json().catch(() => ({}));
// Ollama reports failures as { error } (e.g. 'model "tev1" not found, try pulling it first').
if (!res.ok) throw new Error(data.error || `HTTP ${res.status}`);
} else {
data = await unslothProxy(baseUrl, "/v1/systemone", "POST", payload);
}
if (!data || typeof data.answers !== "object") throw new Error(data?.error || "Decision API returned no answers");
return data.answers;
}
/**
* Run decide() over many states with a small concurrency cap (the model runs locally, so flooding
* it only queues requests). Per-item failures are returned as { error } instead of rejecting all.
*/
export async function decideMany(states, questions, opts = {}) {
const concurrency = Math.max(1, opts.concurrency || 2);
const results = new Array(states.length);
let next = 0;
const worker = async () => {
while (next < states.length) {
const i = next++;
try {
results[i] = { answers: await decide(states[i], questions, opts) };
} catch (e) {
results[i] = { error: e.message };
}
opts.onProgress?.(i, results[i]);
}
};
await Promise.all(Array.from({ length: Math.min(concurrency, states.length) }, worker));
return results;
}
/**
* Connection check for the Settings tab: one yes/no question, timed. The first request after
* Unsloth starts also loads the model (10-20 s), so callers should say that a slow first test is
* normal.
*/
export async function testDecisionConnection(opts = {}) {
const started = performance.now();
const answers = await decide("A cat is sitting on a sofa.", {
test: noul("Is there an animal in the text?"),
}, opts);
return { ms: Math.round(performance.now() - started), yes: answers.test?.noul };
}
// ---- Answer helpers ----
// Threshold on `probabilities` (or noul's value), not `confidence` — per Unsloth's docs, confidence
// only says how peaked the distribution is and its formula differs from Jev's.
/** True when the yes-probability of a noul answer reaches `threshold`. */
export function isYes(answer, threshold = getDecisionSettings().threshold) {
return typeof answer?.noul === "number" && answer.noul >= threshold;
}
/**
* Most likely option of a choice answer. `confident` is false below `threshold` — callers should
* then show the value as a suggestion for the user to confirm rather than apply it automatically.
*/
export function pickChoice(answer, threshold = getDecisionSettings().threshold) {
const value = answer?.choice ?? null;
const probability = value != null ? (answer.probabilities?.[value] ?? 0) : 0;
return { value, probability, confident: value != null && probability >= threshold };
}
/**
* Most likely level of a score answer: { level (index from 0), label, value (expected score,
* fractional), probability, confident }.
*/
export function pickScore(answer, threshold = getDecisionSettings().threshold) {
const probs = answer?.probabilities || {};
let level = null, probability = 0;
for (const [k, p] of Object.entries(probs)) {
if (p > probability) { level = Number(k); probability = p; }
}
return {
level,
label: level != null ? (answer.legend?.[String(level)] ?? null) : null,
value: typeof answer?.score === "number" ? answer.score : null,
probability,
confident: level != null && probability >= threshold,
};
}
+198 -24
View File
@@ -796,7 +796,7 @@ const LANGUAGES = {
helpWf6: "Detail modal: rename, favorite, thumbnail change, badge assignment, tags, memo, AI summary, send to canvas, load in GenerateUI, set as default, delete",
helpWf7: "Badge management: add, rename, delete badges with custom colors shared with the Models tab (⚙ Badge button)",
helpGen1: "Connect/disconnect ComfyUI and check connection status",
helpGen2: "Load workflows from Workflow tab (subgraph-containing workflows are supported, e.g. Flux.2 / Qwen-Image / Z-Image / Mage-Flow / LongCat / Boogu / HiDream E1 / FireRed / Ernie Image templates; only ComfyUI's simplified App format [.app.json / Linear Mode] is not); Reset Workflow button re-reads the current workflow file from disk and reloads the editor; Save button (right end of the tab bar) saves the current workflow — including any edits made in the editor — as a new .json file to the Workflow tab; a filename dialog appears with the current workflow name as the default",
helpGen2: "Load workflows from Workflow tab (subgraph-containing workflows are supported, e.g. Flux.2 / Qwen-Image / Z-Image / Mage-Flow / LongCat / Boogu / HiDream E1 / FireRed / Ernie Image / Ming Image templates; only ComfyUI's simplified App format [.app.json / Linear Mode] is not); Reset Workflow button re-reads the current workflow file from disk and reloads the editor; Save button (right end of the tab bar) saves the current workflow — including any edits made in the editor — as a new .json file to the Workflow tab; a filename dialog appears with the current workflow name as the default",
helpGen25: "Create Catalog button (next to the Style dropdown): generates one preview image per Style selected in the Batch tab's Style sub-tab (using the currently loaded workflow, typically T2I), then saves each result into a chosen ws_style_catalog subfolder with the Style's own name as the filename (existing files with the same name are overwritten); a modal lets you create a new destination folder or pick an existing one — separate folders are useful for grouping catalogs by checkpoint or by purpose; requires at least one Style checked in Batch tab → Style sub-tab first. Catalog button (right next to it): jumps straight to the Gallery tab's Style_Catalog sub-tab to browse the results",
helpGen26: "Create Catalog modal — \"Don't keep a copy in the Output folder\" checkbox: each generation still writes a normal file to ComfyUI's Output folder first (that's simply how the workflow's own Save Image node works), but once its copy has been saved into the catalog folder, checking this deletes that Output-folder original right away — useful for large catalog runs (e.g. all 200+ styles) so the Output folder isn't flooded with near-duplicate files; if a particular style's catalog copy fails to save, its Output file is left alone rather than being deleted",
helpGen27: "Input tab → Image sub-panel — each LoadImage card now has three buttons: Apply (unchanged), Clear (resets that card's pending selection/preview back to empty; doesn't touch the workflow node itself, since LoadImage always needs some valid filename), and Placeholder. A Placeholder Image section — collapsed by default since it's a low-frequency feature, click to expand — sits at the top of the panel with a shared default: Color mode (set Width/Height/Color; generates a solid-color PNG on the fly) or Image mode (drop one default image, reused as-is). Clicking a card's own Placeholder button immediately replaces just that card's image with the configured default — no separate Apply click needed, and with several LoadImage cards each can be replaced individually. The Placeholder config (mode, size, color, default image) is remembered across sessions",
@@ -915,15 +915,21 @@ const LANGUAGES = {
helpPrompt25: "Presets table toolbar — the ★ and B sort buttons (next to the search box) bring favorited / batch-selected presets to the top of the list; both can be enabled together (★ takes priority, B breaks ties), and everything else keeps its original relative order",
helpPrompt26: "Style table toolbar — select exactly one row whose File badge isn't custom.json to enable + Add to this file: it adds a blank row targeting that same file (shown as a badge in the File column even before saving) — the same capability as the Form tab's Style manager (see above); custom.json entries and multi-row selections leave the button disabled, use the regular + New Style button for those instead",
helpSettings1: "UI language and summary language",
helpSettings2: "Theme selection (13 presets) with customization: color overrides, background patterns (stripes, polka dot, checkerboard, SVG tiling), and font selection (16 Google Fonts including Japanese display fonts)",
helpSettings2: "Theme selection (13 presets) with customization: color overrides, background patterns (stripes, polka dot, checkerboard, SVG tiling), and font selection (15 Google Fonts including Japanese display fonts, plus the system default)",
helpSettings3: "Workflow data folder path",
helpSettings4: "ComfyUI connection URL and test",
helpSettings5: "AI Assistant settings have moved out of this tab — the Prompt tab, Tagger tab, and AI TOOL tab each have their own ⚙ Settings button (backend, URL, model, Thinking mode, Max tokens, unload model), saved independently",
helpSettings6: "Default workflow setting",
helpSettings7: "Eagle integration URL and auto-save toggle",
helpSettings8: "Data Management: export core settings, metadata, and prompts to a JSON file, or import to restore. After importing, restart ComfyUI.",
helpSettings15: "Full Backup (ZIP): unlike the JSON export above, also includes folder-based data (AI Skills, Lab Plans, Video Plans, Style presets) and the Tagger database — everything the JSON export can't represent. Restoring overwrites existing files; restart ComfyUI afterward.",
helpSettings16: "Two opt-in checkboxes above the Full Backup buttons additionally include ComfyUI's default workflows folder and the Impact Pack's wildcard folder in the export — both export-only, never restored automatically; copy the ZIP's _external/ folder back by hand if you want them.",
helpSettings15: "Full Backup (ZIP): unlike the JSON export above, also includes folder-based data (AI Skills, Lab Plans, Video Plans, Video Edit projects, Style presets) and the Tagger database — everything the JSON export can't represent. Restoring overwrites existing files; restart ComfyUI afterward.",
helpSettings16: "Two opt-in checkboxes above the Full Backup buttons additionally include ComfyUI's default workflows folder and the Impact Pack's wildcard folder in the export — both export-only, never restored automatically; copy the ZIP's _external/ folder back by hand if you want them. A third checkbox includes the clips and BGM referenced by saved Video Edit projects (projects only store references to files in ComfyUI's input folder) — unlike the other two, these are restored into the input folder, skipping any file that already exists there.",
helpSettings17: "Decision Model: its own section, independent of the AI TOOL / Tagger LLM/VLM backends so it can be used alongside them — backend, URL, model and a default probability threshold (answers below it are treated as suggestions to confirm). Unsloth: Laya models (laya / laya-multilingual / laya-english / laya-typed-decisions); requires Settings → API → Decision API → Serve requests in Unsloth Desktop (otherwise HTTP 404 \"The Decision API is off\") and either UNSLOTH_API_KEY in .env or Keyless API access → \"Chat and inference\". Ollama (0.35+): lists installed models whose capabilities include \"decision\" (tev1, tev1:0.8b, nimble — ollama pull first; ↻ refreshes the list); no key needed. Test sends one yes/no question with the unsaved form values; the first request loads the model.",
helpSettings18: "Video Volume: one slider (0–100%) sets the playback volume of the Video tab preview and the Gallery detail / lightbox / compare video players; the level set here (or changed with any video's own controls) carries over to the next video opened anywhere in the app",
helpSettings19: "Gallery Output Folder: the ComfyUI output folder the Gallery tab browses — leave blank to use ComfyUI's default output directory; Apply takes effect immediately, Default resets it, and the folder currently in use is shown below the field",
helpSettings20: "Default Checkpoint: when \"Force default checkpoint on workflow load\" is on, every checkpoint loader node's model is replaced with the chosen checkpoint whenever a workflow is loaded into GenerateUI (a toast shows when it's applied) — avoids errors or garbled images caused by a model name saved in the workflow that you don't have; Refresh reloads the checkpoint list, Save stores the setting",
helpSettings21: "CivitAI API Key: optional key for authenticated CivitAI access from the Models tab (get it at civitai.com → Account Settings → API Keys); stored in settings.json but left out of the JSON export and Full Backup, and the CIVITAI_API_KEY environment variable takes priority when set",
helpSettings22: "G'MIC-Qt Integration: path to the G'MIC-Qt executable (gmic_qt.exe) used by the Image Edit tab's G'MIC filter (Edit with G'MIC); Save stores it — see the Image Edit help for installing G'MIC-Qt",
helpSettings9: "All sections are collapsible accordion panels",
helpSettings10: "Wildcard Integration: link WFS wildcard dir to ComfyUI-Impact-Pack wildcards (directory junction on Windows / symlink on other OS); existing files migrated automatically",
helpSettings11: "Text Size: one slider (10–28 px) adjusts font size for all prompt and chat textareas at once — Generate UI positive/negative prompts, AI Assistant chat input, Preset positive/negative, Wildcard prompt and file editor, and Metadata prompt full preview; takes effect immediately and saved with Save Settings",
@@ -1016,7 +1022,7 @@ const LANGUAGES = {
helpAi2: "Translation pane: translate between Japanese, English, Chinese, or Free language using Ollama, LM Studio, Lemonade, or Unsloth; ⇄ swap button swaps both language selectors and text content; selections are saved automatically",
helpAi3: "Chat pane: multi-turn conversation with the LLM; Enter to send a message, Shift+Enter for a newline; Clear button resets the conversation history; if the model supports tool calling and asks for an image (\"generate a cat picture\", etc.) it will run generation on the workflow currently loaded in GenerateUI and show the result inline; the \"Allow image generation\" checkbox in the pane header toggles this on/off; if an image is attached (dropped in the TOOLS pane) it is sent to the LLM as vision input and, when generate_image is called, used as the base for image-to-image generation instead of text-to-image; generate_image also accepts optional generation parameters the model can fill in from the request — steps, cfg, sampler_name, scheduler, denoise, seed, width, height, batch_size (number of images produced in that one call), and aspect_ratio (a ratio preset like \"3:2\", applied only if the workflow has a Resolution Selector node feeding width/height — ignored otherwise); anything the model doesn't supply keeps whatever the loaded workflow already has set; models verified to work well for tool calling and vision: Gemma 4:e4b, Qwen 3.5:9b, Ministral 3:3b (Ollama)",
helpAi4: "TOOLS pane — VLM: drop an image (PNG/JPG/WebP) and select a task (Describe Image, Create Prompt, Create Tags), click Run; or select \"Create wildcards\", enter a category name and count to generate plain-text entries one per line; the dropped image is also shared with the Chat pane as its attachment (shown there and used for image-to-image generation) until cleared with the ✕ button",
helpAi6: "Settings pane: choose backend (Ollama / LM Studio / Lemonade / Unsloth), set API URL, test connection, select model, and use the Unload Model button (next to Test connection) to free the loaded model from VRAM/memory — Ollama uses keep_alive:0, LM Studio uses /api/v1/models/unload, Lemonade uses /api/v1/unload; Unsloth has no unload API, so this shows an unsupported message for it; Unsloth always requires an API key even for local access — set UNSLOTH_API_KEY in a .env file in the plugin folder (copy .env.example) and restart ComfyUI, the key is read server-side and never sent to the browser; Unsloth also requires \"Auto model switching (OpenAI API)\" to be turned ON in the Unsloth Desktop app itself (its own Settings screen) — without it, requests fail with HTTP 404 unless a model was already manually loaded in the Unsloth Studio UI; Unsloth reasoning models return their thinking in a separate reasoning_content field rather than inline <think> tags, so a low Max tokens can make the model spend its whole budget reasoning and return an empty reply — increase Max tokens (2048+) if replies come back blank; Generation section — Thinking mode toggle (off by default; Ollama receives a think flag, and regardless of backend any <think>...</think> block is stripped from the output when off, so reasoning models don't clutter Translation/Chat/TOOLS results) and Max tokens (Ollama: num_predict, LM Studio/Lemonade/Unsloth: max_tokens; 0 = unlimited); configure Free language names; Chat Image Generation section lets you enable a dedicated workflow for Chat-triggered text-to-image generation instead of the one currently loaded in GenerateUI; Chat I2I Generation section is the same but for image-to-image generation when an image is attached; settings are shared with the Library panel's A tab via localStorage",
helpAi6: "Settings pane: choose backend (Ollama / LM Studio / Lemonade / Unsloth), set API URL, test connection, select model, and use the Unload Model button (next to Test connection) to free the loaded model from VRAM/memory — Ollama uses keep_alive:0, LM Studio uses /api/v1/models/unload, Lemonade uses /api/v1/unload; Unsloth has no unload API, so this shows an unsupported message for it; Unsloth needs either an API key or Keyless API access — set UNSLOTH_API_KEY in a .env file in the plugin folder (copy .env.example) and restart ComfyUI (the key is read server-side and never sent to the browser), or leave it unset and turn on Keyless API access → \"Chat and inference\" in Unsloth Desktop's Settings → API (requests are then sent without a key; note that setting also opens the API to your private LAN, and keep \"Allow tools\" off); Unsloth also requires \"Auto model switching (OpenAI API)\" to be turned ON in the Unsloth Desktop app itself (its own Settings screen) — without it, requests fail with HTTP 404 unless a model was already manually loaded in the Unsloth Studio UI; Unsloth reasoning models return their thinking in a separate reasoning_content field rather than inline <think> tags, so a low Max tokens can make the model spend its whole budget reasoning and return an empty reply — increase Max tokens (2048+) if replies come back blank; Generation section — Thinking mode toggle (off by default; Ollama receives a think flag, and regardless of backend any <think>...</think> block is stripped from the output when off, so reasoning models don't clutter Translation/Chat/TOOLS results) and Max tokens (Ollama: num_predict, LM Studio/Lemonade/Unsloth: max_tokens; 0 = unlimited); configure Free language names; Chat Image Generation section lets you enable a dedicated workflow for Chat-triggered text-to-image generation instead of the one currently loaded in GenerateUI; Chat I2I Generation section is the same but for image-to-image generation when an image is attached; settings are shared with the Library panel's A tab via localStorage",
helpAi7: "Chat pane — SVG: the LLM can also produce SVG graphics directly as text, with no ComfyUI workflow involved; ask it to draw something in SVG and, if the reply contains SVG code, it is rendered inline as a preview with a \"Save to Gallery\" button that writes it as a .svg file to the Gallery tab's output folder; this does not use tool calling, so it works with any model regardless of tool-calling support",
helpAi8: "Chat pane — Skills: pick a skill from the dropdown above the message list to prepend its instructions to the conversation as a system prompt (not shown in the visible chat log); use the pencil button to open the skill manager and create, edit, or delete .md skill files, stored in user/default/Workflow-Studio/ai_skills/; a file's first line block between --- markers can define name/description shown in the dropdown; two starter skills are included by default: \"SVG Icon Generator\" (decomposes the subject into parts before drawing, for more recognizable SVG output) and \"Skill Creator\" (interviews you conversationally and drafts a new skill file, ready to save); when a reply contains a ```skill code block, a \"Save as new skill\" button appears below it that opens the manager with the draft pre-filled",
helpAi9: "Tips for writing your own skills: give a clear frontmatter name/description so it reads well in the dropdown and tooltip; state the exact output format you want and, if there is a deliverable to extract later (like SVG or another skill file), wrap it in a distinctive fenced code block; write concrete, imperative instructions rather than vague ones - smaller local models follow explicit rules and a short worked example far better than general guidance; state hard constraints (forbidden content, security, length limits) explicitly rather than assuming they are implied; when in doubt, select the \"Skill Creator\" skill in the dropdown and describe what you want - it will interview you and draft the file for you",
@@ -1233,6 +1239,20 @@ const LANGUAGES = {
defaultWorkflowHint: "Set from the workflow detail modal. Auto-loaded when the page opens.",
clear: "Clear",
eagleIntegration: "Eagle Integration (Optional)",
decisionSection: "Decision Model",
decisionHint: "Decision models return calibrated probabilities for typed yes/no, choice and score questions about text (no images). Configured separately from the AI TOOL / Tagger backends so they can be used side by side. Unsloth: Laya via the Decision API (turn on Decision API → Serve requests; uses UNSLOTH_API_KEY from .env, or no key if Keyless API access → \"Chat and inference\" is on). Ollama 0.35+: tev1 / nimble (ollama pull tev1:0.8b etc.; no key).",
decisionBackend: "Backend",
decisionUrl: "URL",
decisionModel: "Model",
decisionModelLaya: "laya (model selected in Unsloth)",
decisionRefreshModels: "Refresh model list",
decisionNoOllamaModels: "No decision models installed in Ollama. Run e.g. \"ollama pull tev1:0.8b\" (812 MB), \"ollama pull tev1\" (4B, 4.5 GB) or \"ollama pull nimble\" (9B, 9.5 GB), then press ↻.",
decisionThreshold: "Default probability threshold",
decisionThresholdHint: "Answers below this probability are treated as suggestions to confirm, not applied automatically.",
decisionTesting: "Testing... (the first request loads the model and can take 10-20 s)",
decisionTestOk: "{ms} ms, test answer yes={yes}",
decisionInvalidUrl: "Invalid URL (http/https only)",
decisionInvalidThreshold: "Threshold must be between 0.5 and 0.99",
eagleUrl: "Eagle URL",
eagleAutoSave: "Auto-save generated images to Eagle",
dataManagement: "Data Management",
@@ -1245,10 +1265,12 @@ const LANGUAGES = {
importError: "Import failed",
exportFullBackup: "Full Backup (ZIP)",
importFullBackup: "Restore Full Backup (ZIP)",
fullBackupHint: "Full Backup (ZIP) additionally includes folder-based data (AI Skills, Lab Plans, Video Plans, Style presets) and the Tagger database — everything the JSON export above can't represent. Restoring overwrites existing files; restart ComfyUI afterward.",
fullBackupHint: "Full Backup (ZIP) additionally includes folder-based data (AI Skills, Lab Plans, Video Plans, Video Edit projects, Style presets) and the Tagger database — everything the JSON export above can't represent. Restoring overwrites existing files; restart ComfyUI afterward.",
fullBackupFilesUnit: "files",
fullBackupIncludeWorkflows: "Include default workflows folder (user/default/workflows)",
fullBackupIncludeWildcard: "Include Impact Pack wildcard folder (if present)",
fullBackupIncludeVideoMedia: "Include media used by Video Edit projects (clips/BGM in the input folder; restored into input without overwriting)",
fullBackupMediaResult: "Video Edit media: {restored} restored, {existing} already present",
fullBackupExternalHint: "These are export-only — restoring a backup never touches them; copy the _external/ folder back manually if needed.",
saveSettings: "Save Settings",
settingsSaved: "Settings saved",
@@ -1361,7 +1383,7 @@ const LANGUAGES = {
videoEditTotalDuration: (time) => `Total: ${time}`,
videoEditPreviewPlay: "▶ Preview",
videoEditPreviewStop: "■ Stop",
videoEditConfirmClear: "Clear the entire edit timeline?",
videoEditConfirmClear: "Clear the entire edit timeline (clips and audio settings)?",
videoSourceAddToEdit: "Add to Edit",
videoEditDuplicate: "Duplicate",
videoEditDelete: "Delete",
@@ -1391,6 +1413,42 @@ const LANGUAGES = {
videoEditProjectSaveFailed: "Failed to save project",
videoEditProjectLoaded: "Project loaded",
videoEditProjectLoadFailed: "Failed to load project",
videoEditTextOverlays: "Text overlays",
videoEditTextAdd: "+ Add text",
videoEditTextEmpty: "No text on this clip",
videoEditTextDefault: "Text",
videoEditTextStart: "Start (s)",
videoEditTextEnd: "End (s)",
videoEditTextSize: "Size (%)",
videoEditTextColor: "Color",
videoEditTextPosition: "Position",
videoEditTextOutline: "Outline",
videoEditTextBackground: "Background",
videoEditTextAnchorLabels: ["↖ Top left", "↑ Top center", "↗ Top right", "← Middle left", "● Center", "→ Middle right", "↙ Bottom left", "↓ Bottom center", "↘ Bottom right"],
videoEditBurningText: "Burning in text...",
videoEditAudioTitle: "Audio",
videoEditKeepAudio: "Keep original audio",
videoEditOriginalVolume: "Original volume",
videoEditBgmSelect: "BGM...",
videoEditBgmRemove: "Remove BGM",
videoEditBgmNone: "No BGM",
videoEditBgmVolume: "BGM volume",
videoEditBgmOffset: "BGM start (s)",
videoEditBgmHint: "Drop an audio file here, or use \"Set as BGM\" on a video in Asset",
videoEditBgmNoAudio: "This video has no audio track",
videoEditBgmUnsupported: "Drop an audio or video file",
videoAssetSetAsBgm: "Set as BGM (audio only)",
videoAssetSetAsBgmDone: (name) => `Set the audio of "${name}" as BGM`,
videoEditCrop: "Crop",
videoEditCropEdit: "Edit crop",
videoEditCropDone: "Done",
videoEditCropCancel: "Cancel",
videoEditCropReset: "Reset",
videoEditCropAspect: "Aspect ratio",
videoEditCropFree: "Free",
videoEditCropNone: "Full frame",
videoEditUndo: "Undo (Ctrl+Z)",
videoEditRedo: "Redo (Ctrl+Y / Ctrl+Shift+Z)",
videoConfirmClearPlan: "Clear the current plan?",
videoPlanPause: "Pause",
videoPlanResume: "Resume",
@@ -1443,7 +1501,7 @@ const LANGUAGES = {
helpVideo8: "Chaining: setting a block's First Image to \"use previous block's last frame\" continues smoothly from the previous clip — if that previous block has its own explicit Last Image, that image is used directly; otherwise the actual last frame of its generated video is extracted automatically.",
helpVideo9: "Project tab (sidebar): lists saved Video Plans while the center Plan subtab is active, or saved Edit projects while the center Edit subtab is active (thumbnail/block or clip count, last updated) — click one to load it, or delete it with the × button; clicking the same tab again collapses the panel.",
helpVideo10: "Highlight color: the Prompt / Duration / First Image / Last Image / Aspect Ratio / Megapixels / Multiple / Noise Seed / Width / Height / Turbo Mode / Strength / Steps / Prompt Enhance labels turn the same highlight color as the GenerateUI Model tab (customizable in the Settings tab) when the loaded workflow actually supports that field — e.g. for a Wan2.2 workflow (no ResolutionSelector node), the Aspect Ratio / Megapixels / Multiple labels stay unhighlighted while Width / Height light up instead.",
helpVideo11: "Edit subtab: a lightweight clip timeline for trimming and concatenating video/image clips outside of generation. Add clips via the Video Source panel's \"Add to Edit\" button or the Asset subtab's \"Send to Edit\" button, then drag blocks to reorder (or use the ◀/▶ toolbar buttons), Duplicate/Delete the selected clip, and Clear to empty the timeline. The trim panel for the selected video clip shows a time ruler, a draggable playhead synced to the Source preview, and draggable in/out handles with time badges — image clips instead get a simple hold-duration field. ▶ Preview plays every ready clip back-to-back (trimmed to its in/out points) in the right-hand result preview; Export builds and runs the actual ComfyUI graph (trim → concatenate if more than one clip → save), tagging the result into the Video Temp group automatically. Exporting more than one clip currently drops audio, since ComfyUI's concatenation node requires uniform audio across all inputs. The export panel's Save/Save As/Load buttons persist the timeline itself (clip order, trim points, which uploaded file each clip refers to) as a JSON project file on the server — Load re-fetches each clip from its saved server reference and re-probes it, so the original files must still exist. These saved projects also show up in the sidebar's Project tab whenever the Edit subtab is active (see helpVideo9).",
helpVideo11: "Edit subtab: a lightweight clip timeline for trimming and concatenating video/image clips outside of generation. Add clips via the Video Source panel's \"Add to Edit\" button or the Asset subtab's \"Send to Edit\" button, then drag blocks to reorder (or use the ◀/▶ toolbar buttons), Duplicate/Delete the selected clip, and Clear to empty the timeline (this also resets the Audio section). The trim panel for the selected video clip shows a time ruler, a draggable playhead synced to the Source preview, and draggable in/out handles with time badges — image clips instead get a simple hold-duration field. ▶ Preview plays every ready clip back-to-back (trimmed to its in/out points) in the right-hand result preview; Export builds and runs the actual ComfyUI graph (trim → concatenate if more than one clip → save), tagging the result into the Video Temp group automatically. The Audio section of the export panel controls the soundtrack: \"Keep original audio\" (with its own volume in dB) keeps each clip's audio aligned to its segment even across multiple clips (image clips contribute silence), and \"BGM...\" lays one audio file over the whole video with its own volume and start offset (a longer BGM is cut to the video length). Each clip can also carry text overlays (\"+ Add text\" under the trim panel: text, start/end seconds within the clip, size as % of frame height, color, 9-position anchor, outline, background box); they are previewed live over the Source/Result panes and burned in on Export as a final pass. The BGM can be chosen with \"BGM...\", by dropping an audio/video file onto the Audio section, or with \"Set as BGM\" on a video in the Asset subtab (its audio track is used). Video clips can also be cropped: \"Edit crop\" shows a draggable/resizable rectangle over the Source preview (optionally locked to 16:9 / 9:16 / 1:1 / 4:3 / 3:4) — \"Done\" keeps it, \"Cancel\" restores the crop from before editing, and switching to another subtab/tab ends editing like Done; the export size becomes the first video clip's cropped size, and other clips of a different size are center-fit to it. ↶/↷ in the timeline toolbar (or Ctrl+Z / Ctrl+Y) undo and redo edits to clips, trims, crops, text and audio settings. The export panel's Save/Save As/Load buttons persist the timeline itself (clip order, trim points, crop, text overlays, audio/BGM settings, and which uploaded file each clip and the BGM refer to) as a JSON project file on the server. The project only stores references to those files in ComfyUI's input folder, so Load re-fetches each clip from its saved server reference and re-probes it — the original files must still exist (Settings → Full Backup can bundle them with \"Include media used by Video Edit projects\"). These saved projects also show up in the sidebar's Project tab whenever the Edit subtab is active (see the Project tab item).",
helpVideo12: "Asset subtab (center panel, next to Plan/Edit): browses every video/image tagged into Video Assets or Video Temp (via the group dropdown, or \"All Video Assets\" for both). Search filters by filename; the ▦/☰ buttons switch between a thumbnail grid and a table (filename/updated/tags). Selecting an item shows its tags/memo, loads it into the center preview, and offers \"Send to Edit\" (queues it as a clip in the Edit subtab) plus, for images only, \"Set as First Image\"/\"Set as Last Image\" (uploads it straight into the currently-selected Plan block, switching to the Plan subtab).",
},
@@ -2237,7 +2295,7 @@ const LANGUAGES = {
helpWf6: "詳細モーダル:名前変更、お気に入り、サムネイル変更、バッジ付与、タグ、メモ、AI要約、キャンバスへ送る、生成UIに読み込み、デフォルト設定、削除",
helpWf7: "バッジ管理:バッジの追加・名前変更・削除、カスタム色設定(Modelsタブと共有)— ⚙ Badgeボタン",
helpGen1: "ComfyUIの接続・切断、接続状態の確認",
helpGen2: "Workflowタブからワークフローを読み込み(サブグラフを含むワークフローにも対応 — Flux.2 / Qwen-Image / Z-Image / Mage-Flow / LongCat / Boogu / HiDream E1 / FireRed / Ernie Image 等のテンプレート含む;ComfyUIの簡略化App形式 [.app.json / Linear Mode] のみ非対応)。Reset Workflowボタンは現在のワークフローファイルをディスクから再読込してエディタをリロード。Saveボタン(タブバー右端)はエディタでの編集を含む現在のワークフローを新しい.jsonファイルとしてWorkflowタブに保存(現在のワークフロー名を初期値としたファイル名ダイアログを表示)",
helpGen2: "Workflowタブからワークフローを読み込み(サブグラフを含むワークフローにも対応 — Flux.2 / Qwen-Image / Z-Image / Mage-Flow / LongCat / Boogu / HiDream E1 / FireRed / Ernie Image / Ming Image 等のテンプレート含む;ComfyUIの簡略化App形式 [.app.json / Linear Mode] のみ非対応)。Reset Workflowボタンは現在のワークフローファイルをディスクから再読込してエディタをリロード。Saveボタン(タブバー右端)はエディタでの編集を含む現在のワークフローを新しい.jsonファイルとしてWorkflowタブに保存(現在のワークフロー名を初期値としたファイル名ダイアログを表示)",
helpGen25: "カタログ作成ボタン(Styleドロップダウンの隣):Batchタブ→Styleサブタブで選択中の各スタイルについて、現在読み込み中のワークフロー(通常はT2I)でプレビュー画像を1枚ずつ生成し、指定したws_style_catalogサブフォルダへスタイル名をファイル名として保存(同名ファイルは上書き)。実行時、新規フォルダ作成または既存フォルダ選択のモーダルが表示される — フォルダを分けることでチェックポイント別・用途別にカタログを整理できる。実行にはBatchタブ→Styleサブタブで最低1つスタイルを選択しておく必要がある。カタログボタン(その隣):Galleryタブ Style_Catalogサブタブへ直接移動して結果を確認できる",
helpGen26: "カタログ作成モーダルの「Outputフォルダにコピーを残さない」チェックボックス:各生成は最初は必ずComfyUIのOutputフォルダへ通常通りファイルを書き出す(ワークフロー自体のSave Imageノードの仕様のため)。このチェックを入れておくと、カタログフォルダへのコピー保存が成功した直後に、そのOutput側の元ファイルを即座に削除する — 200件超の全スタイルなど大規模なカタログ作成でOutputフォルダがほぼ重複したファイルで溢れるのを防げる。特定のスタイルでカタログへのコピー保存自体に失敗した場合、そのOutputファイルは削除されずそのまま残る",
helpGen27: "Inputタブ → Imageサブパネル — 各LoadImageカードにApply(変更なし)・Clear・Placeholderの3ボタンを追加。Clearはそのカードの保留中の選択・プレビューを空に戻す(LoadImageは常に有効なファイル名が必須なため、ワークフローノード自体には触れない)。パネル上部の「Placeholder Image」セクション(使用頻度が低い機能のためデフォルトで折りたたみ、クリックで展開)で共通のデフォルトを設定できる:Colorモード(幅・高さ・色を指定し、その場で単色PNGを生成)またはImageモード(デフォルト画像を1枚ドロップし、そのまま再利用)。各カード自身のPlaceholderボタンを押すと、そのカードの画像だけを設定済みデフォルトへ即座に差し替える(別途Applyクリック不要)— LoadImageが複数ある場合も個別に差し替え可能。Placeholder設定(モード・サイズ・色・デフォルト画像)はセッションをまたいで記憶される",
@@ -2356,15 +2414,21 @@ const LANGUAGES = {
helpPrompt25: "Presetsテーブルのツールバー — 検索ボックス横の★とBソートボタンで、お気に入り/Batch登録済みのプリセットを一覧の先頭に並べ替えます。両方同時にONにでき、その場合は★が優先、Bが次点の基準になり、それ以外は元の相対順序を維持します",
helpPrompt26: "Styleテーブルのツールバー — Fileバッジがcustom.jsonではない行を1件だけ選択すると「+ Add to this file」が有効になります。同じファイルを追加先とする空の行を追加します(保存前からFile列にバッジで表示されます)— Formタブのスタイルマネージャーと同じ機能です(上記参照)。custom.jsonのスタイルや複数選択時はボタンが無効のままです。その場合は通常の「+ New Style」を使ってください",
helpSettings1: "UI言語と要約言語の設定",
helpSettings2: "テーマ選択(13種プリセット)+カスタマイズ:カラー調整、背景パターン(ストライプ、ポルカドット、チェック、SVGタイリング)、フォント選択(日本語デザインフォント含む16種のGoogle Fonts)",
helpSettings2: "テーマ選択(13種プリセット)+カスタマイズ:カラー調整、背景パターン(ストライプ、ポルカドット、チェック、SVGタイリング)、フォント選択(日本語デザインフォントを含むGoogle Fonts 15種+システム既定)",
helpSettings3: "ワークフローデータフォルダのパス変更・リセット",
helpSettings4: "ComfyUI接続URL設定と接続テスト",
helpSettings5: "AI Assistant設定はこのタブから移動しました——Promptタブ・Taggerタブ・AI TOOLタブそれぞれが独自の⚙設定ボタン(バックエンド・URL・モデル・Thinking mode・Max tokens・モデルのアンロード)を持ち、個別に保存されます",
helpSettings6: "デフォルトワークフローの設定・解除",
helpSettings7: "Eagle連携URL、自動保存の有効/無効",
helpSettings8: "データ管理:設定・メタデータ・プロンプトなど主要データをJSONにエクスポート、またはインポートして復元。インポート後はComfyUIを再起動してください。",
helpSettings15: "フルバックアップ (ZIP):上記のJSONエクスポートと異なり、フォルダ形式のデータ(AI Skills、Lab Plan、Video Plan、Styleプリセット)とTaggerデータベースも含みます。復元すると既存ファイルは上書きされます。復元後はComfyUIを再起動してください。",
helpSettings16: "フルバックアップのボタン上にある2つのチェックボックスで、デフォルトワークフローフォルダーとImpact Packのwildcardフォルダーもエクスポートに含められます。どちらもエクスポート専用で自動復元はされません。必要な場合はZIP内の_external/フォルダから手動でコピーしてください。",
helpSettings15: "フルバックアップ (ZIP):上記のJSONエクスポートと異なり、フォルダ形式のデータ(AI Skills、Lab Plan、Video Plan、Video Editプロジェクト、Styleプリセット)とTaggerデータベースも含みます。復元すると既存ファイルは上書きされます。復元後はComfyUIを再起動してください。",
helpSettings16: "フルバックアップのボタン上にある2つのチェックボックスで、デフォルトワークフローフォルダーとImpact Packのwildcardフォルダーもエクスポートに含められます。どちらもエクスポート専用で自動復元はされません。必要な場合はZIP内の_external/フォルダから手動でコピーしてください。3つ目のチェックボックスでは、保存済みVideo Editプロジェクトが参照するクリップ・BGMも含められます(プロジェクトにはComfyUIのinputフォルダ内ファイルへの参照しか保存されないため)。こちらは他の2つと異なり、復元時にinputフォルダへ戻されます(同名ファイルが既にある場合はスキップ)。",
helpSettings17: "意思決定モデル:AI TOOL/TaggerのLLM・VLMと併用できるよう、それらとは独立したセクションで設定します — バックエンド、URL、モデル、既定の確率しきい値(これ未満の回答は確認用の提案として扱う)。Unsloth:Laya系モデル(laya / laya-multilingual / laya-english / laya-typed-decisions)。Unsloth Desktopで Settings → API → Decision API → Serve requests の有効化(無効だとHTTP 404「The Decision API is off」)と、.envのUNSLOTH_API_KEY(またはKeyless API access →「Chat and inference」の有効化)が必要です。Ollama(0.35以降):capabilitiesに「decision」を持つインストール済みモデルを一覧表示します(tev1、tev1:0.8b、nimble — 先にollama pullが必要、↻で一覧を更新)。キーは不要です。テストボタンは保存前のフォーム値でyes/no質問を1問送ります(初回はモデル読み込みに時間がかかります)",
helpSettings18: "Video Volume(動画の音量):スライダー1本(0〜100%)で、Videoタブのプレビューと、Galleryの詳細・ライトボックス・比較表示の動画プレーヤーの再生音量を設定します。ここで設定した音量(または各動画のコントロールで変えた音量)は、アプリ内で次に開く動画にも引き継がれます",
helpSettings19: "Gallery出力フォルダ:Galleryタブで閲覧するComfyUIの出力フォルダ。空欄ならComfyUIの既定の出力ディレクトリを使います。「適用」で即時反映、「デフォルト」で元に戻し、現在使用中のフォルダは入力欄の下に表示されます",
helpSettings20: "デフォルトCheckpoint:「ワークフロー読み込み時にデフォルトCheckpointを強制」をオンにすると、GenerateUIへワークフローを読み込むたびに全てのCheckpointローダーノードのモデルが選択したCheckpointに置き換わります(適用時にトーストで通知)。ワークフローに保存された手元に無いモデル名によるエラーや画像の崩れを防げます。「更新」で一覧を再取得、「保存」で設定を保存します",
helpSettings21: "CivitAI APIキー:ModelsタブからCivitAIへ認証付きでアクセスするための任意のキー(civitai.com → Account Settings → API Keys で取得)。settings.jsonに保存されますが、JSONエクスポートとフルバックアップには含まれません。環境変数CIVITAI_API_KEYが設定されている場合はそちらが優先されます",
helpSettings22: "G'MIC-Qt連携:Image EditタブのG'MICフィルター(Edit with G'MIC)が使うG'MIC-Qt実行ファイル(gmic_qt.exe)のパス。「Save」で保存します。G'MIC-Qtのインストール方法はImage Editのヘルプを参照してください",
helpSettings9: "各セクションはアコーディオン式で折りたたみ可能",
helpSettings10: "Wildcard Integration:WFSのワイルドカードディレクトリをComfyUI-Impact-Packのwildcardsディレクトリへリンク(Windowsはジャンクション、他OSはシンボリックリンク)。既存ファイルは自動で移行されます",
helpSettings11: "テキストサイズ:スライダー1本(10〜28px)でプロンプト・チャット全テキストエリアのフォントサイズを一括調整 — GenerateUI ポジティブ/ネガティブプロンプト、AIアシスタントチャット入力、プリセット ポジティブ/ネガティブ、ワイルドカードプロンプト・ファイルエディタ、メタデータプロンプト全文プレビュー;即時反映・保存で確定",
@@ -2457,7 +2521,7 @@ const LANGUAGES = {
helpAi2: "翻訳ペイン:Ollama、LM Studio、Lemonade、または Unsloth を使って日本語・英語・中国語・カスタム言語(Free)間でテキストを翻訳;⇄ ボタンで言語とテキストを同時入替;選択は自動保存",
helpAi3: "Chat ペイン:LLM とのマルチターン会話;Enter で送信、Shift+Enter で改行;Clear ボタンで会話履歴をリセット;モデルがtool callingに対応している場合、「猫の画像を作って」のように頼むとGenerateUIに読み込み中のワークフローで生成を実行し、結果をチャット内に表示する;ペインヘッダーの「画像生成を許可する」チェックボックスでこの機能のON/OFFを切り替えられる;画像が添付されている場合(TOOLSペインにドロップした画像)はLLMにもビジョン入力として送信され、generate_image呼び出し時はT2Iの代わりにI2I(image-to-image)の元画像として使われる;generate_image呼び出しにはsteps・cfg・sampler_name・scheduler・denoise・seed、width・height、batch_size(その1回の呼び出しで生成する枚数)、aspect_ratio(\"3:2\"のような比率プリセット。ワークフローにwidth/heightを供給するResolution Selectorノードがある場合のみ適用され、それ以外は無視される)といった任意の生成パラメータも指定できる;モデルが指定しなかった項目は読み込み中のワークフローの現在の設定がそのまま使われる;tool callingとビジョンの動作を確認済みのモデル: Gemma 4:e4b、Qwen 3.5:9b、Ministral 3:3b(Ollama)",
helpAi4: "TOOLS ペイン — VLM:画像(PNG/JPG/WebP)をドロップしてタスク(画像解説 / プロンプト作成 / タグ作成)を選んで Run;または「Create wildcards」を選んでカテゴリ名と件数を入力すると1行1エントリのワイルドカードを生成;ドロップした画像はChatペインの添付としても共有され(Chat側にも表示、I2I生成に使用)、✕ボタンでクリアするまで保持される",
helpAi6: "Settings ペイン:バックエンド(Ollama / LM Studio / Lemonade / Unsloth)の選択、APIのURL設定、接続テスト、モデル選択、および「Test connection」の右隣にある「Unload Model」ボタンでロード中のモデルをVRAM/メモリから解放——Ollamaは keep_alive:0、LM Studioは /api/v1/models/unload、Lemonadeは /api/v1/unload を使用;Unslothにはアンロード用APIが存在しないため未対応メッセージが表示される;Unslothはローカルアクセスでも常にAPIキー認証が必須——プラグインフォルダの .env.example を .env にコピーして UNSLOTH_API_KEY を設定し ComfyUI を再起動する(キーはサーバー側でのみ読み込まれ、ブラウザには一切送信されない);Unsloth Desktopアプリ側の設定画面で「モデル自動切り替え (OpenAI API)」もONにする必要がある——OFFのままだと、Unsloth Studio上で事前にモデルを手動ロードしていない限り HTTP 404 で接続に失敗する;Unslothのreasoningモデルは思考内容をインラインの<think>タグではなく別のreasoning_contentフィールドで返すため、Max tokensが少ないとモデルが思考だけでトークンを使い切り、空の返信になることがある——返信が空になる場合はMax tokensを増やす(2048以上を推奨);Generationセクション — Thinking modeトグル(デフォルトOFF;OllamaにはthinkフラグとしてOFF/ONを送信、バックエンドを問わずOFF時は出力中の<think>...</think>ブロックをクライアント側で除去するため、推論モデルの思考過程がTranslation/Chat/TOOLSの結果に混入しない)とMax tokens(Ollama: num_predict、LM Studio/Lemonade/Unsloth: max_tokens;0で無制限);Free言語名の設定;Chat Image Generationセクションでは、ChatからのT2I画像生成にGenerateUIの現在の読み込みワークフローの代わりに専用ワークフローを使うよう設定できる;Chat I2I Generationセクションは同様の設定を画像添付時のI2I生成向けに提供する;設定は localStorage に保存され、Library パネルの A タブと共有される",
helpAi6: "Settings ペイン:バックエンド(Ollama / LM Studio / Lemonade / Unsloth)の選択、APIのURL設定、接続テスト、モデル選択、および「Test connection」の右隣にある「Unload Model」ボタンでロード中のモデルをVRAM/メモリから解放——Ollamaは keep_alive:0、LM Studioは /api/v1/models/unload、Lemonadeは /api/v1/unload を使用;Unslothにはアンロード用APIが存在しないため未対応メッセージが表示される;UnslothはAPIキーかKeyless API accessのどちらかが必要——プラグインフォルダの .env.example を .env にコピーして UNSLOTH_API_KEY を設定し ComfyUI を再起動する(キーはサーバー側でのみ読み込まれ、ブラウザには一切送信されない)か、キーを未設定のままUnsloth DesktopのSettings → APIで Keyless API access →「Chat and inference」をオンにする(キー無しで送信される。この設定は同じLAN内にもAPIを開放するので注意し、「Allow tools」はオフのままにする);Unsloth Desktopアプリ側の設定画面で「モデル自動切り替え (OpenAI API)」もONにする必要がある——OFFのままだと、Unsloth Studio上で事前にモデルを手動ロードしていない限り HTTP 404 で接続に失敗する;Unslothのreasoningモデルは思考内容をインラインの<think>タグではなく別のreasoning_contentフィールドで返すため、Max tokensが少ないとモデルが思考だけでトークンを使い切り、空の返信になることがある——返信が空になる場合はMax tokensを増やす(2048以上を推奨);Generationセクション — Thinking modeトグル(デフォルトOFF;OllamaにはthinkフラグとしてOFF/ONを送信、バックエンドを問わずOFF時は出力中の<think>...</think>ブロックをクライアント側で除去するため、推論モデルの思考過程がTranslation/Chat/TOOLSの結果に混入しない)とMax tokens(Ollama: num_predict、LM Studio/Lemonade/Unsloth: max_tokens;0で無制限);Free言語名の設定;Chat Image Generationセクションでは、ChatからのT2I画像生成にGenerateUIの現在の読み込みワークフローの代わりに専用ワークフローを使うよう設定できる;Chat I2I Generationセクションは同様の設定を画像添付時のI2I生成向けに提供する;設定は localStorage に保存され、Library パネルの A タブと共有される",
helpAi7: "Chat ペイン — SVG:LLMはComfyUIワークフローを一切使わず、SVGグラフィックをテキストとして直接生成することもできる;「SVGで〜を描いて」のように頼み、応答にSVGコードが含まれていれば自動でインラインプレビュー表示され、「Galleryに保存」ボタンでGalleryタブの出力フォルダに.svgファイルとして保存できる;tool calling(function calling)を使用しないため、tool calling非対応のモデルでも動作する",
helpAi8: "Chat ペイン — スキル:メッセージ一覧上部のドロップダウンでスキルを選ぶと、その指示内容が system prompt として会話の先頭に付加される(表示上のチャット履歴には含まれない);鉛筆ボタンでスキル管理パネルを開き、.mdスキルファイルを作成・編集・削除できる(保存先: user/default/Workflow-Studio/ai_skills/);ファイル先頭の---で囲まれた部分にname/descriptionを書くとドロップダウンの表示に反映される;初期状態で2つのスキルを同梱: 「SVG Icon Generator」(描く前に対象を部位分解することで認識しやすいSVGを生成)と「Skill Creator」(対話形式で質問しながら新しいスキルファイルを下書きする);応答に```skillコードブロックが含まれる場合、メッセージ下部に「スキルとして保存」ボタンが表示され、内容が入った状態で管理パネルが開く",
helpAi9: "スキル作成のヒント:ドロップダウンやツールチップで分かりやすいよう、frontmatterのname/descriptionは明確に書く;期待する出力形式を具体的に指示し、SVGや別のスキルファイルのように後で抽出したい成果物がある場合は専用のコードフェンスで囲むよう指示する;曖昧な指示より具体的で命令形の指示の方がローカルの小型モデルには効きやすく、短い実例を1つ添えるとさらに効果的;禁止事項(不適切な内容、セキュリティ、文字数制限など)は暗黙に期待せず明示的に書く;迷ったらドロップダウンで「Skill Creator」スキルを選び、作りたいスキルの概要を伝えれば対話形式で下書きしてくれる",
@@ -2674,6 +2738,20 @@ const LANGUAGES = {
defaultWorkflowHint: "ワークフロー詳細モーダルから設定してください。ページ起動時に自動読み込みされます。",
clear: "クリア",
eagleIntegration: "Eagle 連携 (オプション)",
decisionSection: "意思決定モデル",
decisionHint: "意思決定モデルは、テキストに対するyes/no・選択・段階評価の質問に確率つきで回答します(画像は不可)。AI TOOL/TaggerのLLM・VLMと併用できるよう、それらとは別に設定します。Unsloth:Decision API経由のLaya(Decision API → Serve requests を有効化。APIキーは.envのUNSLOTH_API_KEY、未設定ならKeyless API access →「Chat and inference」がオンなら動作)。Ollama 0.35以降:tev1/nimble(ollama pull tev1:0.8b 等で取得。キー不要)。",
decisionBackend: "バックエンド",
decisionUrl: "URL",
decisionModel: "モデル",
decisionModelLaya: "laya(Unsloth側で選択中のモデル)",
decisionRefreshModels: "モデル一覧を更新",
decisionNoOllamaModels: "Ollamaに意思決定モデルがありません。「ollama pull tev1:0.8b」(812MB)、「ollama pull tev1」(4B・4.5GB)、「ollama pull nimble」(9B・9.5GB)などで取得してから ↻ を押してください。",
decisionThreshold: "既定の確率しきい値",
decisionThresholdHint: "この確率未満の回答は自動適用せず、確認用の提案として扱います。",
decisionTesting: "テスト中...(初回はモデル読み込みで10〜20秒かかります)",
decisionTestOk: "{ms} ms、テスト回答 yes={yes}",
decisionInvalidUrl: "URLが不正です(http/httpsのみ)",
decisionInvalidThreshold: "しきい値は0.5〜0.99で指定してください",
eagleUrl: "Eagle URL",
eagleAutoSave: "生成画像をEagleに自動保存",
dataManagement: "データ管理",
@@ -2686,10 +2764,12 @@ const LANGUAGES = {
importError: "インポート失敗",
exportFullBackup: "フルバックアップ (ZIP)",
importFullBackup: "フルバックアップを復元 (ZIP)",
fullBackupHint: "フルバックアップ(ZIP)は上記のJSONエクスポートでは表現できないフォルダ形式のデータ(AI Skills、Lab Plan、Video Plan、Styleプリセット)とTaggerデータベースも含みます。復元すると既存ファイルは上書きされます。復元後はComfyUIを再起動してください。",
fullBackupHint: "フルバックアップ(ZIP)は上記のJSONエクスポートでは表現できないフォルダ形式のデータ(AI Skills、Lab Plan、Video Plan、Video Editプロジェクト、Styleプリセット)とTaggerデータベースも含みます。復元すると既存ファイルは上書きされます。復元後はComfyUIを再起動してください。",
fullBackupFilesUnit: "件",
fullBackupIncludeWorkflows: "デフォルトワークフローフォルダーを含める (user/default/workflows)",
fullBackupIncludeWildcard: "Impact Packのwildcardフォルダーを含める(存在する場合)",
fullBackupIncludeVideoMedia: "Video Editプロジェクトの素材を含める(inputフォルダ内のクリップ・BGM。復元時はinputへ戻し、既存ファイルは上書きしません)",
fullBackupMediaResult: "Video Edit素材: {restored}件復元、{existing}件は既存",
fullBackupExternalHint: "これらはエクスポート専用です。復元時には一切触れません。必要な場合はZIP内の_external/フォルダから手動でコピーしてください。",
saveSettings: "設定を保存",
settingsSaved: "設定を保存しました",
@@ -2802,7 +2882,7 @@ const LANGUAGES = {
videoEditTotalDuration: (time) => `合計: ${time}`,
videoEditPreviewPlay: "▶ プレビュー",
videoEditPreviewStop: "■ 停止",
videoEditConfirmClear: "編集タイムラインを全てクリアしますか?",
videoEditConfirmClear: "編集タイムライン(クリップとオーディオ設定)を全てクリアしますか?",
videoSourceAddToEdit: "Editに追加",
videoEditDuplicate: "複製",
videoEditDelete: "削除",
@@ -2832,6 +2912,42 @@ const LANGUAGES = {
videoEditProjectSaveFailed: "プロジェクトの保存に失敗しました",
videoEditProjectLoaded: "プロジェクトを読み込みました",
videoEditProjectLoadFailed: "プロジェクトの読み込みに失敗しました",
videoEditTextOverlays: "テキストオーバーレイ",
videoEditTextAdd: "+ テキスト追加",
videoEditTextEmpty: "このクリップにテキストはありません",
videoEditTextDefault: "テキスト",
videoEditTextStart: "開始(秒)",
videoEditTextEnd: "終了(秒)",
videoEditTextSize: "サイズ(%)",
videoEditTextColor: "色",
videoEditTextPosition: "位置",
videoEditTextOutline: "縁取り",
videoEditTextBackground: "背景",
videoEditTextAnchorLabels: ["↖ 左上", "↑ 上中央", "↗ 右上", "← 左中央", "● 中央", "→ 右中央", "↙ 左下", "↓ 下中央", "↘ 右下"],
videoEditBurningText: "テキストを焼き込み中...",
videoEditAudioTitle: "オーディオ",
videoEditKeepAudio: "元の音声を残す",
videoEditOriginalVolume: "元音声の音量",
videoEditBgmSelect: "BGM選択...",
videoEditBgmRemove: "BGMを外す",
videoEditBgmNone: "BGMなし",
videoEditBgmVolume: "BGM音量",
videoEditBgmOffset: "BGM開始位置(秒)",
videoEditBgmHint: "音声ファイルをここへドロップ、またはAssetの動画で「BGMに設定」",
videoEditBgmNoAudio: "この動画には音声トラックがありません",
videoEditBgmUnsupported: "音声または動画ファイルをドロップしてください",
videoAssetSetAsBgm: "BGMに設定(音声のみ)",
videoAssetSetAsBgmDone: (name) => `「${name}」の音声をBGMに設定しました`,
videoEditCrop: "クロップ",
videoEditCropEdit: "クロップ編集",
videoEditCropDone: "完了",
videoEditCropCancel: "キャンセル",
videoEditCropReset: "リセット",
videoEditCropAspect: "アスペクト比",
videoEditCropFree: "自由",
videoEditCropNone: "全体",
videoEditUndo: "元に戻す (Ctrl+Z)",
videoEditRedo: "やり直し (Ctrl+Y / Ctrl+Shift+Z)",
videoConfirmClearPlan: "現在のPlanをクリアしますか?",
videoPlanPause: "一時停止",
videoPlanResume: "再開",
@@ -2884,7 +3000,7 @@ const LANGUAGES = {
helpVideo8: "チェイン機能: ブロックのFirst Imageを「前ブロックの最終フレームを使用」に設定すると、前のクリップから連続した映像になります。前ブロックに明示的なLast Imageが設定されていればそれがそのまま使われ、無ければ生成された動画から実際の最終フレームが自動抽出されます。",
helpVideo9: "Projectタブ(左サイドバー): 中央がPlanサブタブのときは保存済みPlan一覧、Editサブタブのときは保存済みEditプロジェクト一覧(サムネイル/ブロック数またはクリップ数・更新日時)を表示します。クリックで読み込み、×ボタンで削除できます。同じタブをもう一度押すとパネルを閉じます。",
helpVideo10: "ハイライト表示: 読み込んだワークフローが実際にサポートしているフィールドに応じて、Prompt/Duration/First Image/Last Image/Aspect Ratio/Megapixels/Multiple/Noise Seed/Width/Height/Turbo Mode/Strength/Steps/Prompt Enhanceのラベルが、GenerateUI ModelタブのハイライトとSettingsタブで共通設定できる色に変わります。例えばWan2.2ワークフロー(ResolutionSelectorノードが無い)ではAspect Ratio/Megapixels/Multipleはハイライトされない代わりに、Width/Heightがハイライトされます。",
helpVideo11: "Editサブタブ: 生成とは独立して動画/画像クリップをトリム・結合できる簡易タイムラインです。Video Sourceパネルの「Editに追加」ボタン、またはAssetサブタブの「Editへ送る」ボタンでクリップを追加し、ブロックをドラッグ(または◀/▶ツールバーボタン)で並べ替え、選択中クリップをDuplicate/Delete、Clearでタイムラインを空にできます。選択中の動画クリップのトリムパネルには時間目盛り・Sourceプレビューと同期するドラッグ可能な再生ヘッド・時刻バッジ付きのドラッグ可能なイン/アウトハンドルが表示されます(画像クリップは代わりにシンプルな保持時間の入力欄)。「▶ プレビュー」は準備済みの全クリップをイン/アウト点でトリムして右側の結果プレビューで連続再生し、「書き出し」は実際のComfyUIグラフ(トリム→2クリップ以上ならConcatenate→保存)を構築・実行して、結果をVideo Tempグループへ自動登録します。複数クリップの書き出しは現状音声が失われます(ComfyUIの結合ノードが全入力で音声形式の統一を要求するため)。書き出しパネルのSave/Save As/Loadボタンは、タイムラインそのもの(クリップ順序・トリム区間・各クリップが参照するアップロード済みファイル)をサーバー上にJSONプロジェクトとして保存します。Loadは保存されたサーバー参照から各クリップを再取得・再プローブするため、元のファイルが存在している必要があります。保存済みプロジェクトは、中央がEditサブタブのときサイドバーのProjectタブにも一覧表示されます(helpVideo9参照)。",
helpVideo11: "Editサブタブ: 生成とは独立して動画/画像クリップをトリム・結合できる簡易タイムラインです。Video Sourceパネルの「Editに追加」ボタン、またはAssetサブタブの「Editへ送る」ボタンでクリップを追加し、ブロックをドラッグ(または◀/▶ツールバーボタン)で並べ替え、選択中クリップをDuplicate/Delete、Clearでタイムラインを空にできます(オーディオ欄も初期状態に戻ります)。選択中の動画クリップのトリムパネルには時間目盛り・Sourceプレビューと同期するドラッグ可能な再生ヘッド・時刻バッジ付きのドラッグ可能なイン/アウトハンドルが表示されます(画像クリップは代わりにシンプルな保持時間の入力欄)。「▶ プレビュー」は準備済みの全クリップをイン/アウト点でトリムして右側の結果プレビューで連続再生し、「書き出し」は実際のComfyUIグラフ(トリム→2クリップ以上ならConcatenate→保存)を構築・実行して、結果をVideo Tempグループへ自動登録します。書き出しパネルのオーディオ欄でサウンドトラックを設定できます。「元の音声を残す」(音量dB調整付き)は複数クリップでも各クリップの音声をその区間に揃えて保持し(画像クリップ区間は無音)、「BGM選択...」で1本の音声ファイルを動画全体に重ねられます(音量・開始位置を指定可、動画より長いBGMは動画の長さでカット)。また各クリップにはテキストオーバーレイを付けられます(トリムパネル下の「+ テキスト追加」: 文字・クリップ内の開始/終了秒・サイズ(フレーム高さに対する%)・色・9方向の位置・縁取り・背景)。Source/Resultプレビュー上でリアルタイムに確認でき、書き出し時に最終工程として焼き込まれます。BGMは「BGM選択...」ボタン、オーディオ欄への音声/動画ファイルのドラッグ&ドロップ、またはAssetサブタブの動画の「BGMに設定」(その動画の音声トラックを使用)で指定できます。動画クリップはクロップもできます。「クロップ編集」でSourceプレビュー上にドラッグ/リサイズ可能な枠が表示されます(16:9 / 9:16 / 1:1 / 4:3 / 3:4 に固定も可)。「完了」で確定、「キャンセル」で編集前の状態に戻り、他のサブタブ/タブへ移動すると「完了」と同じ扱いで編集を終了します。書き出しサイズは先頭の動画クリップのクロップ後サイズになり、サイズの異なる他のクリップは中央基準で合わせられます。タイムラインツールバーの ↶/↷(または Ctrl+Z / Ctrl+Y)で、クリップ・トリム・クロップ・テキスト・オーディオ設定の編集を元に戻す/やり直しできます。書き出しパネルのSave/Save As/Loadボタンは、タイムラインそのもの(クリップ順序・トリム区間・クロップ・テキストオーバーレイ・音声/BGM設定、および各クリップとBGMが参照するアップロード済みファイル)をサーバー上にJSONプロジェクトとして保存します。プロジェクトに保存されるのはComfyUIのinputフォルダ内ファイルへの参照だけなので、Loadは保存されたサーバー参照から各クリップを再取得・再プローブし、元のファイルが存在している必要があります(Settingsのフルバックアップで「Video Editプロジェクトの素材を含める」を使うと素材ごと保存できます)。保存済みプロジェクトは、中央がEditサブタブのときサイドバーのProjectタブにも一覧表示されます(Projectタブの項目を参照)。",
helpVideo12: "Assetサブタブ(中央パネル、Plan/Editの隣): グループ切替(または「All Video Assets」で両方まとめて)でVideo AssetsまたはVideo Tempに登録された動画/画像を一覧表示します。検索欄はファイル名でフィルタし、▦/☰ボタンでサムネイルグリッドとテーブル表示(ファイル名・更新日時・タグ)を切り替えられます。項目を選択するとタグ・メモが表示され中央プレビューに読み込まれ、「Editへ送る」(Editサブタブのクリップとして追加)に加え、画像の場合のみ「First Imageに設定」「Last Imageに設定」(選択中のPlanブロックへ直接アップロードし、Planサブタブへ切り替え)が使えます。",
},
@@ -3678,7 +3794,7 @@ const LANGUAGES = {
helpWf6: "详情弹窗:重命名、收藏、缩略图更换、徽章分配、标签、备注、AI摘要、发送到画布、加载到生成UI、设为默认、删除",
helpWf7: "徽章管理:添加、重命名、删除徽章并自定义颜色(与Models标签共享)— ⚙ Badge按钮",
helpGen1: "连接/断开ComfyUI,检查连接状态",
helpGen2: "从工作流标签页加载工作流(支持含子图的工作流,例如 Flux.2 / Qwen-Image / Z-Image / Mage-Flow / LongCat / Boogu / HiDream E1 / FireRed / Ernie Image 等模板;仅不支持ComfyUI的简化App格式 [.app.json / Linear Mode])。Reset Workflow按钮从磁盘重新读取当前工作流文件并重新加载编辑器。Save按钮(标签栏右端)将当前工作流(包括在编辑器中所做的编辑)保存为新的.json文件到工作流标签页;会显示以当前工作流名称为默认值的文件名对话框",
helpGen2: "从工作流标签页加载工作流(支持含子图的工作流,例如 Flux.2 / Qwen-Image / Z-Image / Mage-Flow / LongCat / Boogu / HiDream E1 / FireRed / Ernie Image / Ming Image 等模板;仅不支持ComfyUI的简化App格式 [.app.json / Linear Mode])。Reset Workflow按钮从磁盘重新读取当前工作流文件并重新加载编辑器。Save按钮(标签栏右端)将当前工作流(包括在编辑器中所做的编辑)保存为新的.json文件到工作流标签页;会显示以当前工作流名称为默认值的文件名对话框",
helpGen25: "创建目录按钮(Style下拉框旁):为Batch标签→Style子标签中选中的每个风格,使用当前加载的工作流(通常为T2I)各生成一张预览图,并以风格名称为文件名保存到指定的ws_style_catalog子文件夹(同名文件将被覆盖);会弹出对话框让你新建目标文件夹或选择已有文件夹 — 使用多个文件夹便于按checkpoint或用途分类整理目录;执行前需先在Batch标签→Style子标签中至少勾选一个风格。目录按钮(紧邻其右):直接跳转到Gallery标签的Style_Catalog子标签查看结果",
helpGen26: "创建目录对话框的「不在Output文件夹中保留副本」复选框:每次生成仍会先按工作流自身Save Image节点的机制,正常写入一份文件到ComfyUI的Output文件夹;勾选此项后,一旦该图像的副本已成功保存到目录文件夹,就会立即删除Output文件夹中的原始文件 — 适合大批量创建目录(例如全部200多个风格)时,避免Output文件夹被大量近似重复的文件淹没;若某个风格的目录副本保存失败,其Output文件不会被删除,会保留原样",
helpGen27: "Input标签页 → Image子面板 — 每张LoadImage卡片现在有三个按钮:Apply(不变)、Clear(将该卡片待处理的选择/预览重置为空;不会改动工作流节点本身,因为LoadImage始终需要一个有效的文件名)和Placeholder。面板顶部的「Placeholder Image」区域(属于低频功能,默认折叠,点击可展开)用于配置共享的默认值:Color模式(设置宽度/高度/颜色,即时生成纯色PNG)或Image模式(拖入一张默认图片,之后原样复用)。点击某张卡片自己的Placeholder按钮,会立即只替换该卡片的图片为已配置的默认值——无需再点Apply;当有多张LoadImage卡片时,每张都可以单独替换。Placeholder配置(模式、尺寸、颜色、默认图片)会跨会话记住",
@@ -3797,15 +3913,21 @@ const LANGUAGES = {
helpPrompt25: "Presets表格工具栏 — 搜索框旁边的★和B排序按钮,可将收藏/已加入Batch的预设排到列表顶部;两者可同时启用(★优先,B为次要条件),其余项保持原有的相对顺序",
helpPrompt26: "Style表格工具栏——只选中一行且其File徽标不是custom.json时,「+ Add to this file」按钮才会启用:会添加一个以该文件为目标的空白行(保存前File列就会显示该文件的徽标)——与Form标签中Style管理器的同一功能(见上文)。custom.json条目或多选时按钮保持禁用,此时请改用常规的「+ New Style」按钮",
helpSettings1: "UI语言和摘要语言设置",
helpSettings2: "主题选择(13种预设)+ 自定义:颜色调整、背景图案(条纹、波点、棋盘格、SVG平铺)、字体选择(含日文显示字体在内的16种Google Fonts)",
helpSettings2: "主题选择(13种预设)+ 自定义:颜色调整、背景图案(条纹、波点、棋盘格、SVG平铺)、字体选择(含日文显示字体在内的15种Google Fonts及系统默认字体)",
helpSettings3: "工作流数据文件夹路径更改/重置",
helpSettings4: "ComfyUI连接URL设置和测试",
helpSettings5: "AI助手设置已从此标签页移出——Prompt标签页、Tagger标签页、AI TOOL标签页现在各自拥有独立的⚙设置按钮(后端、URL、模型、Thinking mode、Max tokens、卸载模型),分别单独保存",
helpSettings6: "默认工作流设置/取消",
helpSettings7: "Eagle集成URL、自动保存开关",
helpSettings8: "数据管理:将设置、元数据、提示词等主要数据导出为JSON文件,或导入以恢复。导入后请重启ComfyUI。",
helpSettings15: "完整备份 (ZIP):与上方的JSON导出不同,还包含文件夹型数据(AI Skills、Lab Plan、Video Plan、Style预设)以及Tagger数据库。恢复会覆盖现有文件,恢复后请重启ComfyUI。",
helpSettings16: "完整备份按钮上方的两个复选框可将ComfyUI默认工作流文件夹和Impact Pack的wildcard文件夹一并导出。两者均仅用于导出,不会自动恢复。如需要,请从ZIP中的_external/文件夹手动复制回去。",
helpSettings15: "完整备份 (ZIP):与上方的JSON导出不同,还包含文件夹型数据(AI Skills、Lab Plan、Video Plan、Video Edit项目、Style预设)以及Tagger数据库。恢复会覆盖现有文件,恢复后请重启ComfyUI。",
helpSettings16: "完整备份按钮上方的两个复选框可将ComfyUI默认工作流文件夹和Impact Pack的wildcard文件夹一并导出。两者均仅用于导出,不会自动恢复。如需要,请从ZIP中的_external/文件夹手动复制回去。第三个复选框可包含已保存的Video Edit项目所引用的片段和BGM(项目中只保存对ComfyUI input文件夹内文件的引用)。与另外两项不同,这些文件在恢复时会放回input文件夹(已存在同名文件时跳过)。",
helpSettings17: "决策模型:独立的设置区块,与 AI TOOL / Tagger 的 LLM/VLM 后端分开,可同时使用 — 后端、URL、模型以及默认概率阈值(低于该值的回答作为待确认的建议)。Unsloth:Laya 系列模型(laya / laya-multilingual / laya-english / laya-typed-decisions);需要在 Unsloth Desktop 中启用 Settings → API → Decision API → Serve requests(否则返回 HTTP 404“The Decision API is off”),并在 .env 中设置 UNSLOTH_API_KEY(或开启 Keyless API access →「Chat and inference」)。Ollama(0.35+):列出 capabilities 含「decision」的已安装模型(tev1、tev1:0.8b、nimble — 需先 ollama pull,↻ 刷新列表);无需密钥。测试按钮使用未保存的表单值发送一个是/否问题(首次请求需加载模型)。",
helpSettings18: "Video Volume(视频音量):用一个滑块(0–100%)设置 Video 标签预览以及 Gallery 详情 / 灯箱 / 对比视图中视频播放器的音量;此处设置的音量(或在任一视频自身控件中调整的音量)会沿用到应用内下一个打开的视频",
helpSettings19: "Gallery 输出文件夹:Gallery 标签浏览的 ComfyUI 输出文件夹——留空则使用 ComfyUI 默认输出目录;「应用」即时生效,「默认」恢复默认,当前使用的文件夹显示在输入框下方",
helpSettings20: "默认 Checkpoint:开启「加载工作流时强制使用默认 Checkpoint」后,每次将工作流加载到 GenerateUI 时,所有 Checkpoint 加载节点的模型都会替换为所选的 Checkpoint(应用时会弹出提示)——可避免工作流中保存的本地不存在的模型名导致的错误或图像异常;「刷新」重新获取列表,「保存」保存设置",
helpSettings21: "CivitAI API 密钥:用于从 Models 标签以认证方式访问 CivitAI 的可选密钥(在 civitai.com → Account Settings → API Keys 获取);保存在 settings.json 中,但不包含在 JSON 导出和完整备份中;若设置了环境变量 CIVITAI_API_KEY,则优先使用该变量",
helpSettings22: "G'MIC-Qt 集成:Image Edit 标签的 G'MIC 滤镜(Edit with G'MIC)所使用的 G'MIC-Qt 可执行文件(gmic_qt.exe)路径;点击「Save」保存——G'MIC-Qt 的安装方法请参阅 Image Edit 帮助",
helpSettings9: "各部分可折叠展开",
helpSettings10: "Wildcard Integration:将WFS通配符目录链接到ComfyUI-Impact-Pack的wildcards目录(Windows为目录联结,其他OS为符号链接);现有文件自动迁移",
helpSettings11: "文字大小:单个滑块(10–28 px)可一次性调整所有提示词和聊天文本框的字体大小 — GenerateUI正/负向提示词、AI助手聊天输入、预设正/负向提示、通配符提示和文件编辑器、元数据提示词全文预览;即时生效并通过「保存设置」保存",
@@ -3898,7 +4020,7 @@ const LANGUAGES = {
helpAi2: "翻译面板:使用 Ollama、LM Studio、Lemonade 或 Unsloth 在日语、英语、中文或自定义语言(Free)之间翻译;⇄ 按钮同时互换语言和文本内容;选择自动保存",
helpAi3: "Chat 面板:与 LLM 进行多轮对话;Enter 发送消息,Shift+Enter 换行;Clear 按钮重置对话历史;如果模型支持 tool calling,当你要求生成图像(例如\"生成一张猫的图片\")时,会使用 GenerateUI 中当前加载的工作流执行生成,并在对话中显示结果;面板标题栏的\"允许生成图像\"复选框可开关此功能;如果附加了图像(在 TOOLS 面板中拖放的图像),该图像会作为视觉输入发送给 LLM,调用 generate_image 时会以该图像为基础执行图生图(I2I),而不是文生图;generate_image 还接受可选的生成参数——steps、cfg、sampler_name、scheduler、denoise、seed、width、height、batch_size(该次调用生成的图像数量),以及 aspect_ratio(形如\"3:2\"的比例预设,仅当工作流中存在向 width/height 供值的 Resolution Selector 节点时才会生效,否则忽略);模型未提供的参数将保留当前已加载工作流的设置;已验证可正常支持 tool calling 与视觉功能的模型:Gemma 4:e4b、Qwen 3.5:9b、Ministral 3:3b(Ollama)",
helpAi4: "TOOLS 面板 — VLM:拖放图像(PNG/JPG/WebP)并选择任务(图像描述 / 提示词生成 / 标签生成),点击 Run;或选择「Create wildcards」,输入类别名称和数量生成纯文本通配符条目(每行一条);拖放的图像也会作为 Chat 面板的附件共享(在该处显示并用于图生图生成),直到点击 ✕ 按钮清除",
helpAi6: "Settings 面板:选择后端(Ollama / LM Studio / Lemonade / Unsloth)、设置 API URL、测试连接、选择模型,以及使用「Test connection」右侧的「Unload Model」按钮将已加载的模型从显存/内存中卸载——Ollama 使用 keep_alive:0,LM Studio 使用 /api/v1/models/unload,Lemonade 使用 /api/v1/unload;Unsloth 没有卸载 API,因此会显示不支持的提示;Unsloth 即使在本地访问时也始终需要 API 密钥认证——将插件文件夹中的 .env.example 复制为 .env 并设置 UNSLOTH_API_KEY,然后重启 ComfyUI(密钥仅在服务器端读取,绝不会发送到浏览器);还需要在 Unsloth Desktop 应用自身的设置界面中开启「Auto model switching (OpenAI API)」——若保持关闭,除非已在 Unsloth Studio 中手动加载过模型,否则请求会以 HTTP 404 失败;Unsloth 的推理模型会将思考内容放在单独的 reasoning_content 字段中,而不是内联的 <think> 标签里,因此 Max tokens 太小会导致模型把全部配额都用在思考上,返回空白回复——如果回复为空,请调高 Max tokens(建议 2048 以上);Generation 部分 — Thinking mode 开关(默认关闭;Ollama会收到think标志,无论使用哪个后端,关闭时都会从输出中去除<think>...</think>思考块,避免推理模型的思考过程混入Translation/Chat/TOOLS的结果)以及 Max tokens(Ollama: num_predict,LM Studio/Lemonade/Unsloth: max_tokens;0为不限);配置 Free 语言名称;Chat 图像生成部分可让 Chat 触发的文生图使用专用工作流,而不是 GenerateUI 中当前加载的工作流;Chat I2I 生成部分提供相同设置,用于附加图像时的图生图生成;设置保存至 localStorage,并与 Library 面板的 A 标签共享",
helpAi6: "Settings 面板:选择后端(Ollama / LM Studio / Lemonade / Unsloth)、设置 API URL、测试连接、选择模型,以及使用「Test connection」右侧的「Unload Model」按钮将已加载的模型从显存/内存中卸载——Ollama 使用 keep_alive:0,LM Studio 使用 /api/v1/models/unload,Lemonade 使用 /api/v1/unload;Unsloth 没有卸载 API,因此会显示不支持的提示;Unsloth 需要 API 密钥或 Keyless API access 之一——将插件文件夹中的 .env.example 复制为 .env 并设置 UNSLOTH_API_KEY,然后重启 ComfyUI(密钥仅在服务器端读取,绝不会发送到浏览器);或者不设置密钥,在 Unsloth Desktop 的 Settings → API 中开启 Keyless API access →「Chat and inference」(请求将不带密钥发送;该设置也会向局域网开放 API,请注意,并保持「Allow tools」关闭);还需要在 Unsloth Desktop 应用自身的设置界面中开启「Auto model switching (OpenAI API)」——若保持关闭,除非已在 Unsloth Studio 中手动加载过模型,否则请求会以 HTTP 404 失败;Unsloth 的推理模型会将思考内容放在单独的 reasoning_content 字段中,而不是内联的 <think> 标签里,因此 Max tokens 太小会导致模型把全部配额都用在思考上,返回空白回复——如果回复为空,请调高 Max tokens(建议 2048 以上);Generation 部分 — Thinking mode 开关(默认关闭;Ollama会收到think标志,无论使用哪个后端,关闭时都会从输出中去除<think>...</think>思考块,避免推理模型的思考过程混入Translation/Chat/TOOLS的结果)以及 Max tokens(Ollama: num_predict,LM Studio/Lemonade/Unsloth: max_tokens;0为不限);配置 Free 语言名称;Chat 图像生成部分可让 Chat 触发的文生图使用专用工作流,而不是 GenerateUI 中当前加载的工作流;Chat I2I 生成部分提供相同设置,用于附加图像时的图生图生成;设置保存至 localStorage,并与 Library 面板的 A 标签共享",
helpAi7: "Chat 面板 — SVG:LLM 还可以直接以文本形式生成 SVG 图形,完全不涉及 ComfyUI 工作流;只需要求它用 SVG 绘制内容,如果回复中包含 SVG 代码,会自动渲染为内嵌预览,并提供\"保存到图库\"按钮,将其保存为 .svg 文件到 Gallery 标签页的输出文件夹;该功能不使用 tool calling(函数调用),因此不支持 tool calling 的模型也可以使用",
helpAi8: "Chat 面板 — 技能:在消息列表上方的下拉菜单中选择一个技能,其内容会作为 system prompt 添加到对话开头(不会显示在可见的聊天记录中);点击铅笔按钮打开技能管理面板,可创建、编辑、删除 .md 技能文件(保存路径:user/default/Workflow-Studio/ai_skills/);文件开头 --- 之间的 name/description 字段会显示在下拉菜单中;默认内置两个技能:「SVG Icon Generator」(在绘制前先将对象拆解为多个部件,以生成更易识别的 SVG)和「Skill Creator」(通过对话向你提问并起草新的技能文件);当回复中包含```skill代码块时,消息下方会出现「保存为技能」按钮,点击后会打开管理面板并预填内容",
helpAi9: "编写技能的小贴士:为 frontmatter 的 name/description 写清楚的内容,便于在下拉菜单和提示中阅读;明确说明期望的输出格式,如果有需要后续提取的成果物(如 SVG 或另一个技能文件),指示它用专门的代码围栏包裹;具体、祈使句式的指令比笼统的说明对本地小模型更有效,附上一个简短示例效果更佳;明确写出硬性约束(禁止内容、安全性、长度限制等),不要指望模型会自行推断;如果不确定怎么写,可以在下拉菜单中选择「Skill Creator」技能并描述你想要的技能,它会通过提问帮你起草",
@@ -4115,6 +4237,20 @@ const LANGUAGES = {
defaultWorkflowHint: "在工作流详情弹窗中设置。页面打开时自动加载。",
clear: "清除",
eagleIntegration: "Eagle 集成 (可选)",
decisionSection: "决策模型",
decisionHint: "决策模型针对文本的是/否、选择和评分问题返回校准后的概率(不支持图像)。与 AI TOOL / Tagger 后端分开配置,可同时使用。Unsloth:通过 Decision API 使用 Laya(需启用 Decision API → Serve requests;API 密钥使用 .env 中的 UNSLOTH_API_KEY,未设置则需开启 Keyless API access →「Chat and inference」)。Ollama 0.35+:tev1 / nimble(用 ollama pull tev1:0.8b 等获取;无需密钥)。",
decisionBackend: "后端",
decisionUrl: "URL",
decisionModel: "模型",
decisionModelLaya: "laya(Unsloth 中选定的模型)",
decisionRefreshModels: "刷新模型列表",
decisionNoOllamaModels: "Ollama 中没有决策模型。请先运行「ollama pull tev1:0.8b」(812MB)、「ollama pull tev1」(4B,4.5GB)或「ollama pull nimble」(9B,9.5GB),然后点击 ↻。",
decisionThreshold: "默认概率阈值",
decisionThresholdHint: "低于此概率的回答不会自动应用,而是作为待确认的建议。",
decisionTesting: "测试中...(首次请求需加载模型,约 10-20 秒)",
decisionTestOk: "{ms} ms,测试回答 yes={yes}",
decisionInvalidUrl: "URL 无效(仅限 http/https)",
decisionInvalidThreshold: "阈值必须在 0.5 到 0.99 之间",
eagleUrl: "Eagle URL",
eagleAutoSave: "自动保存生成图片到Eagle",
dataManagement: "数据管理",
@@ -4127,10 +4263,12 @@ const LANGUAGES = {
importError: "导入失败",
exportFullBackup: "完整备份 (ZIP)",
importFullBackup: "恢复完整备份 (ZIP)",
fullBackupHint: "完整备份(ZIP)还包含上方JSON导出无法表示的文件夹型数据(AI Skills、Lab Plan、Video Plan、Style预设)以及Tagger数据库。恢复会覆盖现有文件,恢复后请重启ComfyUI。",
fullBackupHint: "完整备份(ZIP)还包含上方JSON导出无法表示的文件夹型数据(AI Skills、Lab Plan、Video Plan、Video Edit项目、Style预设)以及Tagger数据库。恢复会覆盖现有文件,恢复后请重启ComfyUI。",
fullBackupFilesUnit: "个文件",
fullBackupIncludeWorkflows: "包含默认工作流文件夹 (user/default/workflows)",
fullBackupIncludeWildcard: "包含Impact Pack的wildcard文件夹(如存在)",
fullBackupIncludeVideoMedia: "包含Video Edit项目使用的素材(input文件夹中的片段/BGM;恢复到input,不覆盖已有文件)",
fullBackupMediaResult: "Video Edit素材:已恢复 {restored} 个,{existing} 个已存在",
fullBackupExternalHint: "这些仅用于导出,恢复备份时不会处理它们。如需要,请从ZIP中的_external/文件夹手动复制回去。",
saveSettings: "保存设置",
settingsSaved: "设置已保存",
@@ -4243,7 +4381,7 @@ const LANGUAGES = {
videoEditTotalDuration: (time) => `总计: ${time}`,
videoEditPreviewPlay: "▶ 预览",
videoEditPreviewStop: "■ 停止",
videoEditConfirmClear: "确定清空整个编辑时间线吗?",
videoEditConfirmClear: "确定清空整个编辑时间线(片段和音频设置)吗?",
videoSourceAddToEdit: "添加到Edit",
videoEditDuplicate: "复制",
videoEditDelete: "删除",
@@ -4273,6 +4411,42 @@ const LANGUAGES = {
videoEditProjectSaveFailed: "项目保存失败",
videoEditProjectLoaded: "项目已加载",
videoEditProjectLoadFailed: "项目加载失败",
videoEditTextOverlays: "文字叠加",
videoEditTextAdd: "+ 添加文字",
videoEditTextEmpty: "此片段没有文字",
videoEditTextDefault: "文字",
videoEditTextStart: "开始(秒)",
videoEditTextEnd: "结束(秒)",
videoEditTextSize: "大小(%)",
videoEditTextColor: "颜色",
videoEditTextPosition: "位置",
videoEditTextOutline: "描边",
videoEditTextBackground: "背景",
videoEditTextAnchorLabels: ["↖ 左上", "↑ 上中", "↗ 右上", "← 左中", "● 中央", "→ 右中", "↙ 左下", "↓ 下中", "↘ 右下"],
videoEditBurningText: "正在烧录文字...",
videoEditAudioTitle: "音频",
videoEditKeepAudio: "保留原始音频",
videoEditOriginalVolume: "原始音量",
videoEditBgmSelect: "选择BGM...",
videoEditBgmRemove: "移除BGM",
videoEditBgmNone: "无BGM",
videoEditBgmVolume: "BGM音量",
videoEditBgmOffset: "BGM起始(秒)",
videoEditBgmHint: "将音频文件拖放到此处,或在 Asset 中对视频使用“设为BGM”",
videoEditBgmNoAudio: "该视频没有音轨",
videoEditBgmUnsupported: "请拖放音频或视频文件",
videoAssetSetAsBgm: "设为BGM(仅音频)",
videoAssetSetAsBgmDone: (name) => `已将“${name}”的音频设为BGM`,
videoEditCrop: "裁切画面",
videoEditCropEdit: "编辑裁切",
videoEditCropDone: "完成",
videoEditCropCancel: "取消",
videoEditCropReset: "重置",
videoEditCropAspect: "宽高比",
videoEditCropFree: "自由",
videoEditCropNone: "完整画面",
videoEditUndo: "撤销 (Ctrl+Z)",
videoEditRedo: "重做 (Ctrl+Y / Ctrl+Shift+Z)",
videoConfirmClearPlan: "要清空当前 Plan 吗?",
videoPlanPause: "暂停",
videoPlanResume: "继续",
@@ -4325,7 +4499,7 @@ const LANGUAGES = {
helpVideo8: "链接功能:将某个区块的 First Image 设置为“使用前一个区块的最后一帧”,即可与上一个片段流畅衔接——如果前一个区块已明确设置了 Last Image,将直接使用该图片;否则会自动从其生成的视频中提取实际的最后一帧。",
helpVideo9: "Project标签页(左侧边栏):中央为 Plan 子标签时显示已保存的 Plan 列表,为 Edit 子标签时显示已保存的 Edit 项目列表(缩略图/区块数或片段数、最后更新时间),点击即可加载,或用 × 按钮删除;再次点击同一标签会收起面板。",
helpVideo10: "高亮颜色:当所加载的工作流实际支持该字段时,Prompt / Duration / First Image / Last Image / Aspect Ratio / Megapixels / Multiple / Noise Seed / Width / Height / Turbo Mode / Strength / Steps / Prompt Enhance 标签会变为与 GenerateUI Model 标签页相同的高亮颜色(可在 Settings 标签页自定义)——例如 Wan2.2 工作流(没有 ResolutionSelector 节点)中,Aspect Ratio / Megapixels / Multiple 不会高亮,取而代之的是 Width / Height 会高亮。",
helpVideo11: "Edit 子标签:独立于生成流程的轻量剪辑时间线,用于裁剪和拼接视频/图片片段。通过 Video Source 面板的「Add to Edit」按钮,或 Asset 子标签的「Send to Edit」按钮添加片段,拖动区块(或使用 ◀/▶ 工具栏按钮)重新排序,对选中片段执行 Duplicate/Delete,Clear 可清空时间线。选中视频片段的裁剪面板会显示时间刻度尺、与 Source 预览同步的可拖动播放头,以及带时间标签的可拖动入/出点手柄(图片片段则只有一个简单的保持时长输入框)。「▶ Preview」会按入/出点依次裁剪并在右侧结果预览区连续播放所有就绪片段;「Export」会构建并运行实际的 ComfyUI 图(裁剪→若多于一个片段则拼接→保存),并自动把结果标记进 Video Temp 分组。目前导出多个片段会丢失音频,因为 ComfyUI 的拼接节点要求所有输入的音频格式一致。导出面板的 Save/Save As/Load 按钮会把时间线本身(片段顺序、裁剪区间、每个片段引用的已上传文件)作为 JSON 项目保存到服务器;Load 会根据保存的服务器引用重新获取并重新探测每个片段,因此原始文件必须仍然存在。已保存的项目在中央为 Edit 子标签时也会显示在侧边栏的 Project 标签中(参见 helpVideo9)。",
helpVideo11: "Edit 子标签:独立于生成流程的轻量剪辑时间线,用于裁剪和拼接视频/图片片段。通过 Video Source 面板的「Add to Edit」按钮,或 Asset 子标签的「Send to Edit」按钮添加片段,拖动区块(或使用 ◀/▶ 工具栏按钮)重新排序,对选中片段执行 Duplicate/Delete,Clear 可清空时间线(音频区域也会恢复初始状态)。选中视频片段的裁剪面板会显示时间刻度尺、与 Source 预览同步的可拖动播放头,以及带时间标签的可拖动入/出点手柄(图片片段则只有一个简单的保持时长输入框)。「▶ Preview」会按入/出点依次裁剪并在右侧结果预览区连续播放所有就绪片段;「Export」会构建并运行实际的 ComfyUI 图(裁剪→若多于一个片段则拼接→保存),并自动把结果标记进 Video Temp 分组。导出面板的音频区域用于设置音轨:“保留原始音频”(可调节音量 dB)在多个片段拼接时也会让各片段的音频与其区间对齐(图片片段区间为静音);“选择BGM...”可将一个音频文件叠加到整个视频上(可设置音量与起始位置,比视频长的 BGM 会按视频长度截断)。此外,每个片段都可以添加文字叠加(裁剪面板下方的“+ 添加文字”:文字、片段内开始/结束秒数、大小(占画面高度的%)、颜色、9 方位位置、描边、背景),可在 Source/Result 预览上实时查看,并在导出时作为最后一步烧录进视频。BGM 可通过“选择BGM...”按钮、将音频/视频文件拖放到音频区域,或在 Asset 子标签中对视频使用“设为BGM”(使用该视频的音轨)来指定。视频片段还可以裁切画面:点击“编辑裁切”会在 Source 预览上显示可拖动/缩放的裁切框(可锁定为 16:9 / 9:16 / 1:1 / 4:3 / 3:4),“完成”确认,“取消”恢复到编辑前状态,切换到其他子标签/标签页时按“完成”处理并结束编辑;导出尺寸为第一个视频片段裁切后的尺寸,尺寸不同的其他片段会以中心为基准适配。时间轴工具栏的 ↶/↷(或 Ctrl+Z / Ctrl+Y)可撤销/重做对片段、裁剪、裁切、文字和音频设置的编辑。导出面板的 Save/Save As/Load 按钮会把时间线本身(片段顺序、裁剪区间、裁切、文字叠加、音频/BGM 设置,以及每个片段和 BGM 引用的已上传文件)作为 JSON 项目保存到服务器。项目中只保存对 ComfyUI input 文件夹内文件的引用,因此 Load 会根据保存的服务器引用重新获取并重新探测每个片段,原始文件必须仍然存在(在 Settings 的完整备份中勾选“包含Video Edit项目使用的素材”即可连同素材一起备份)。已保存的项目在中央为 Edit 子标签时也会显示在侧边栏的 Project 标签中(参见 Project 标签一项)。",
helpVideo12: "Asset 子标签(中央面板,位于 Plan/Edit 旁):浏览标记到 Video Assets 或 Video Temp 分组的所有视频/图片(通过分组下拉框选择,或选择「All Video Assets」同时浏览两者)。搜索框按文件名过滤,▦/☰ 按钮可在缩略图网格和表格视图(文件名/更新时间/标签)之间切换。选中一项会显示其标签/备注并加载到中央预览,此外还提供「Send to Edit」(作为片段加入 Edit 子标签);仅图片还额外提供「Set as First Image」「Set as Last Image」(直接上传到当前选中的 Plan 区块,并切换到 Plan 子标签)。",
},
};
+20 -14
View File
@@ -359,18 +359,16 @@ function extractPromptsFromNodeSet(nodes, links) {
continue;
}
if (!isTextEncoderNode(type)) continue;
const text = n.widgets_values?.[0];
if (text && typeof text === "string") {
textMap.set(n.id, text);
} else if (Array.isArray(n.inputs)) {
const textInput = n.inputs.find(inp => inp.name === "text" || inp.name === "text_g" || inp.name === "prompt");
if (textInput?.link != null) {
const originId = linkOrigin.get(textInput.link);
const originSlot = linkSlot.get(textInput.link) ?? 0;
const text2 = resolveLinkedTextInNodeSet(nodeMap, linkOrigin, linkSlot, originId, originSlot);
if (text2) textMap.set(n.id, text2);
}
}
// A text widget converted to a wired input keeps its last typed value in widgets_values[0]
// (stale — never sent to the backend), so the linked source must win over it.
const textInput = Array.isArray(n.inputs)
? n.inputs.find(inp => inp.name === "text" || inp.name === "text_g" || inp.name === "prompt")
: null;
const linkedText = textInput?.link != null && linkOrigin.has(textInput.link)
? resolveLinkedTextInNodeSet(nodeMap, linkOrigin, linkSlot, linkOrigin.get(textInput.link), linkSlot.get(textInput.link) ?? 0)
: null;
const text = linkedText || n.widgets_values?.[0];
if (text && typeof text === "string") textMap.set(n.id, text);
}
const pos = new Set(), neg = new Set();
let foundSampler = false;
@@ -381,11 +379,13 @@ function extractPromptsFromNodeSet(nodes, links) {
// Guider node (CFGGuider/DualCFGGuider/BasicGuider) via its "guider" input. Scan that
// node's inputs instead so the loop below can find the positive/negative-role slots.
let inputsToScan = n.inputs;
let viaBasicGuider = false;
const guiderInput = n.inputs.find(inp => inp.name === "guider");
if (guiderInput?.link != null) {
const guiderId = linkOrigin.get(guiderInput.link);
const guiderNode = guiderId != null ? nodeMap.get(guiderId) : null;
if (Array.isArray(guiderNode?.inputs)) inputsToScan = guiderNode.inputs;
viaBasicGuider = guiderNode?.type === "BasicGuider";
}
for (const inp of inputsToScan) {
if (!inp || inp.link == null) continue;
@@ -393,7 +393,9 @@ function extractPromptsFromNodeSet(nodes, links) {
if (originId == null) continue;
const name = inp.name ?? "";
// DualCFGGuider (HiDream E1): cond1 carries the positive-derived conditioning.
const isPos = name === "positive" || name === "cond1" || name.startsWith("positive");
// BasicGuider (Flux/Ming Image etc., no negative): its single "conditioning" input is positive.
const isPos = name === "positive" || name === "cond1" || name.startsWith("positive")
|| (viaBasicGuider && name === "conditioning");
const isNeg = name === "negative" || name.startsWith("negative");
if (!isPos && !isNeg) continue;
// TextEncodeMageFlowEdit / TextEncodeBooguEdit — one node carries both prompt &
@@ -562,14 +564,18 @@ function extractPromptsAPI(wf) {
// Guider node (CFGGuider/DualCFGGuider/BasicGuider) via its "guider" input. Scan that
// node's inputs instead so the loop below can find the positive/negative-role keys.
let inputsToScan = n.inputs ?? {};
let viaBasicGuider = false;
if (Array.isArray(n.inputs?.guider)) {
const guiderNode = wf[String(n.inputs.guider[0])];
if (guiderNode?.inputs) inputsToScan = guiderNode.inputs;
viaBasicGuider = guiderNode?.class_type === "BasicGuider";
}
for (const [key, val] of Object.entries(inputsToScan)) {
if (!Array.isArray(val)) continue;
// DualCFGGuider (HiDream E1): cond1 carries the positive-derived conditioning.
const isPos = key === "positive" || key === "cond1" || key.startsWith("positive");
// BasicGuider (Flux/Ming Image etc., no negative): its single "conditioning" input is positive.
const isPos = key === "positive" || key === "cond1" || key.startsWith("positive")
|| (viaBasicGuider && key === "conditioning");
const isNeg = key === "negative" || key.startsWith("negative");
if (!isPos && !isNeg) continue;
// TextEncodeMageFlowEdit / TextEncodeBooguEdit — one node carries both prompt &
+135 -3
View File
@@ -6,7 +6,8 @@ import { showToast } from "./app.js";
import { comfyUI } from "./comfyui-client.js";
import { t, getLang, getSummaryLang, setLang, setSummaryLang, getLanguageOptions, getSummaryLanguageOptions } from "./i18n.js";
import { getSettings, readJsonStorage } from "./util.js";
import { getSettings, readJsonStorage, escapeHtml } from "./util.js";
import { DECISION_BACKENDS, getDecisionSettings, saveDecisionSettings, testDecisionConnection, listDecisionModels } from "./decision-client.js";
const SETTINGS_KEY = "wfm_settings";
@@ -455,6 +456,7 @@ export async function initSettingsTab() {
const uiLang = getLang();
const summaryLang = getSummaryLang();
const decisionSettings = getDecisionSettings();
container.innerHTML = `
<h2 style="font-size:18px;margin-bottom:20px;">${t("settingsTitle")}</h2>
@@ -803,6 +805,49 @@ export async function initSettingsTab() {
</div>
</details>
<!-- Decision Model (Laya) — independent of the AI TOOL / Tagger LLM/VLM backends -->
<details class="wfm-settings-section">
<summary class="wfm-settings-summary">${t("decisionSection")}</summary>
<small style="color:var(--wfm-text-secondary);font-size:11px;display:block;margin-bottom:8px;">
${t("decisionHint")}
</small>
<div class="wfm-form-group">
<label>${t("decisionBackend")}</label>
<select class="wfm-select" id="wfm-settings-decision-backend">
${Object.entries(DECISION_BACKENDS).map(([b, def]) => `<option value="${b}" ${decisionSettings.backend === b ? "selected" : ""}>${def.label}</option>`).join("")}
</select>
</div>
<div class="wfm-form-group">
<label>${t("decisionUrl")}</label>
<div style="display:flex;gap:8px;">
<input type="text" class="wfm-input" id="wfm-settings-decision-url"
value="${decisionSettings.baseUrl}"
placeholder="${DECISION_BACKENDS[decisionSettings.backend].defaultUrl}">
<button class="wfm-btn" id="wfm-settings-decision-test">${t("test")}</button>
</div>
<div id="wfm-settings-decision-status" style="font-size:12px;margin-top:4px;"></div>
</div>
<div class="wfm-form-group">
<label>${t("decisionModel")}</label>
<div style="display:flex;gap:8px;">
<select class="wfm-select" id="wfm-settings-decision-model" style="flex:1;">
<option value="${escapeHtml(decisionSettings.model)}" selected>${escapeHtml(decisionSettings.model)}</option>
</select>
<button class="wfm-btn" id="wfm-settings-decision-refresh" title="${t("decisionRefreshModels")}">&#8635;</button>
</div>
<div id="wfm-settings-decision-model-hint" style="font-size:11px;color:var(--wfm-text-secondary);margin-top:4px;"></div>
</div>
<div class="wfm-form-group">
<label>${t("decisionThreshold")}</label>
<input type="number" class="wfm-input" id="wfm-settings-decision-threshold"
value="${decisionSettings.threshold}" min="0.5" max="0.99" step="0.05" style="width:100px;">
<small style="color:var(--wfm-text-secondary);font-size:11px;display:block;margin-top:4px;">
${t("decisionThresholdHint")}
</small>
</div>
<button class="wfm-btn wfm-btn-primary wfm-btn-sm" id="wfm-settings-decision-save">${t("save")}</button>
</details>
<!-- G'MIC Integration -->
<details class="wfm-settings-section">
<summary class="wfm-settings-summary">G'MIC-Qt Integration</summary>
@@ -842,6 +887,10 @@ export async function initSettingsTab() {
<input type="checkbox" id="wfm-settings-backup-include-wildcard">
${t("fullBackupIncludeWildcard")}
</label>
<label class="wfm-lab-checkbox-label" style="font-size:12px;">
<input type="checkbox" id="wfm-settings-backup-include-video-media">
${t("fullBackupIncludeVideoMedia")}
</label>
<span style="font-size:11px;color:var(--wfm-text-secondary);">${t("fullBackupExternalHint")}</span>
</div>
<div style="display:flex;gap:8px;flex-wrap:wrap;">
@@ -1237,6 +1286,82 @@ export async function initSettingsTab() {
}
});
// --- Decision Model (Laya) ---
const readDecisionForm = () => {
const baseUrl = document.getElementById("wfm-settings-decision-url")?.value.trim() || "";
const threshold = parseFloat(document.getElementById("wfm-settings-decision-threshold")?.value);
let urlOk = false;
try { urlOk = ["http:", "https:"].includes(new URL(baseUrl).protocol); } catch {}
if (!urlOk) throw new Error(t("decisionInvalidUrl"));
if (!(threshold >= 0.5 && threshold <= 0.99)) throw new Error(t("decisionInvalidThreshold"));
return {
backend: document.getElementById("wfm-settings-decision-backend")?.value || "unsloth",
baseUrl,
model: document.getElementById("wfm-settings-decision-model")?.value || "",
threshold,
};
};
// Unsloth: fixed Laya names. Ollama: installed models whose capabilities include "decision";
// none installed (or Ollama unreachable) → keep the preferred name selectable and show why.
const refreshDecisionModels = async (preferred) => {
const backend = document.getElementById("wfm-settings-decision-backend")?.value || "unsloth";
const baseUrl = document.getElementById("wfm-settings-decision-url")?.value.trim() || "";
const select = document.getElementById("wfm-settings-decision-model");
const hint = document.getElementById("wfm-settings-decision-model-hint");
if (!select) return;
const def = DECISION_BACKENDS[backend];
const current = preferred ?? select.value;
let models = [];
let hintText = "";
try {
({ models } = await listDecisionModels(backend, baseUrl));
if (backend === "ollama" && models.length === 0) hintText = t("decisionNoOllamaModels");
} catch (err) {
hintText = `${t("failedConnect")}: ${err.message}`;
}
const list = models.length ? [...models] : [...def.models];
if (current && !list.includes(current) && (backend === "ollama") === !/^laya/.test(current)) list.unshift(current);
const selected = list.includes(current) ? current : (list.includes(def.defaultModel) ? def.defaultModel : list[0]);
select.innerHTML = list.map((m) => `<option value="${escapeHtml(m)}" ${m === selected ? "selected" : ""}>${m === "laya" ? t("decisionModelLaya") : escapeHtml(m)}</option>`).join("");
if (hint) hint.textContent = hintText;
};
refreshDecisionModels(decisionSettings.model);
document.getElementById("wfm-settings-decision-refresh")?.addEventListener("click", () => refreshDecisionModels());
document.getElementById("wfm-settings-decision-backend")?.addEventListener("change", (e) => {
const urlInput = document.getElementById("wfm-settings-decision-url");
const def = DECISION_BACKENDS[e.target.value]?.defaultUrl;
if (urlInput && def) { urlInput.value = def; urlInput.placeholder = def; }
refreshDecisionModels(DECISION_BACKENDS[e.target.value]?.defaultModel);
});
document.getElementById("wfm-settings-decision-save")?.addEventListener("click", () => {
try {
saveDecisionSettings(readDecisionForm());
showToast(t("settingsSaved"), "success");
} catch (err) {
showToast(err.message, "error");
}
});
// Tests the values currently in the form (saved or not), same as the Eagle test button.
document.getElementById("wfm-settings-decision-test")?.addEventListener("click", async () => {
const statusEl = document.getElementById("wfm-settings-decision-status");
let form;
try { form = readDecisionForm(); } catch (err) {
statusEl.textContent = err.message;
statusEl.style.color = "var(--wfm-danger)";
return;
}
statusEl.textContent = t("decisionTesting");
statusEl.style.color = "var(--wfm-text-secondary)";
try {
const { ms, yes } = await testDecisionConnection({ backend: form.backend, baseUrl: form.baseUrl, model: form.model });
statusEl.textContent = `${t("connectedCheck")} ${t("decisionTestOk").replace("{ms}", ms).replace("{yes}", typeof yes === "number" ? yes.toFixed(3) : "?")}`;
statusEl.style.color = "var(--wfm-success)";
} catch (err) {
statusEl.textContent = `${t("failedConnect")}: ${err.message}`;
statusEl.style.color = "var(--wfm-danger)";
}
});
// Test Eagle URL (via server proxy to avoid CORS)
document.getElementById("wfm-settings-test-eagle")?.addEventListener("click", async () => {
const urlInput = document.getElementById("wfm-settings-eagle-url");
@@ -1491,7 +1616,8 @@ export async function initSettingsTab() {
try {
const includeWorkflows = document.getElementById("wfm-settings-backup-include-workflows")?.checked ? "1" : "0";
const includeWildcard = document.getElementById("wfm-settings-backup-include-wildcard")?.checked ? "1" : "0";
const res = await fetch(`/api/wfm/settings/export-full?include_workflows=${includeWorkflows}&include_wildcard=${includeWildcard}`);
const includeVideoMedia = document.getElementById("wfm-settings-backup-include-video-media")?.checked ? "1" : "0";
const res = await fetch(`/api/wfm/settings/export-full?include_workflows=${includeWorkflows}&include_wildcard=${includeWildcard}&include_video_media=${includeVideoMedia}`);
if (!res.ok) throw new Error(await res.text());
const blob = await res.blob();
const url = URL.createObjectURL(blob);
@@ -1522,7 +1648,13 @@ export async function initSettingsTab() {
});
const result = await res.json();
if (!res.ok) throw new Error(result.error || res.statusText);
statusEl.textContent = `${t("importSuccess")}: ${result.extracted?.length ?? 0} ${t("fullBackupFilesUnit")}`;
let msg = `${t("importSuccess")}: ${result.extracted?.length ?? 0} ${t("fullBackupFilesUnit")}`;
const restoredMedia = result.media_restored?.length ?? 0;
const existingMedia = result.media_existing?.length ?? 0;
if (restoredMedia || existingMedia) {
msg += ` / ${t("fullBackupMediaResult").replace("{restored}", restoredMedia).replace("{existing}", existingMedia)}`;
}
statusEl.textContent = msg;
statusEl.style.color = "var(--wfm-success, #4caf50)";
} catch (e) {
statusEl.textContent = t("importError") + ": " + e.message;
+19 -1
View File
@@ -12,7 +12,7 @@ import { VIDEO_GROUP, VTEMP_GROUP, ensureVideoGroup, isVideoFile } from "./galle
import { setSourcePreview } from "./video-preview.js";
import { showToast } from "./app.js";
import { t } from "./i18n.js";
import { addClipFromFile } from "./video-edit-tab.js";
import { addClipFromFile, setBgmFromFile } from "./video-edit-tab.js";
import { setBlockImageFromFile } from "./video-plan-tab.js";
// Sentinel for the "All Video Assets" option — not a real backend group, since
@@ -242,6 +242,7 @@ function _renderDetail(img) {
<textarea id="wfm-video-asset-memo" class="wfm-textarea" rows="3"></textarea>
<button type="button" class="wfm-btn wfm-btn-sm" id="wfm-video-asset-memo-save" style="margin-top:6px;">Save Memo</button>
<button type="button" class="wfm-btn wfm-btn-sm" id="wfm-video-asset-send-to-edit" style="width:100%;margin-top:12px;">${t("videoEditSendToEdit")}</button>
<button type="button" class="wfm-btn wfm-btn-sm" id="wfm-video-asset-set-bgm" style="width:100%;margin-top:6px;display:${isVideoFile(img) ? "block" : "none"};">${t("videoAssetSetAsBgm")}</button>
<div id="wfm-video-asset-set-plan-image-row" style="display:${isVideoFile(img) ? "none" : "flex"};gap:6px;margin-top:6px;">
<button type="button" class="wfm-btn wfm-btn-sm" id="wfm-video-asset-set-first" style="flex:1;">${t("videoAssetSetAsFirst")}</button>
<button type="button" class="wfm-btn wfm-btn-sm" id="wfm-video-asset-set-last" style="flex:1;">${t("videoAssetSetAsLast")}</button>
@@ -287,6 +288,23 @@ function _renderDetail(img) {
}
});
// Gallery lists no audio files, so a video asset's own audio track is the
// Asset-side way to pick a BGM (LoadAudio decodes audio from video files).
panel.querySelector("#wfm-video-asset-set-bgm")?.addEventListener("click", async () => {
try {
const res = await fetch(`/wfm/gallery/image/serve?path=${encodeURIComponent(img.path)}`);
if (!res.ok) throw new Error(String(res.status));
const blob = await res.blob();
const file = new File([blob], img.filename, { type: blob.type || "video/mp4" });
if (await setBgmFromFile(file, img.filename)) {
document.querySelector('.wfm-video-center-panel .wfm-video-subtab-btn[data-video-subtab="edit"]')?.click();
showToast(t("videoAssetSetAsBgmDone", img.filename), "success");
}
} catch (err) {
showToast(t("errorWithMsg", err.message), "error");
}
});
const setAsPlanImage = async (which) => {
try {
const res = await fetch(`/wfm/gallery/image/serve?path=${encodeURIComponent(img.path)}`);
File diff suppressed because it is too large Load Diff
+15 -2
View File
@@ -18,8 +18,8 @@
*/
const _PANES = {
source: { video: "wfm-video-source-preview-video", img: "wfm-video-source-preview-img", placeholder: "wfm-video-source-preview-placeholder" },
result: { video: "wfm-video-preview-video", img: "wfm-video-preview-img", placeholder: "wfm-video-preview-placeholder" },
source: { frame: "wfm-video-source-preview-frame", video: "wfm-video-source-preview-video", img: "wfm-video-source-preview-img", placeholder: "wfm-video-source-preview-placeholder" },
result: { frame: "wfm-video-preview-frame", video: "wfm-video-preview-video", img: "wfm-video-preview-img", placeholder: "wfm-video-preview-placeholder" },
};
const _objectUrls = { source: null, result: null };
@@ -111,3 +111,16 @@ export function getAllPreviewVideoElements() {
export function getResultPreviewVideoElement() {
return document.getElementById(_PANES.result.video);
}
// The frame/<video>/<img> elements of one pane ("source" | "result") — used by
// video-edit-tab.js to position its text-overlay preview layer over whichever
// media element is currently showing in that pane.
export function getPreviewPaneElements(pane) {
const ids = _PANES[pane];
if (!ids) return null;
return {
frame: document.getElementById(ids.frame),
video: document.getElementById(ids.video),
img: document.getElementById(ids.img),
};
}
+10
View File
@@ -132,6 +132,16 @@ function _applyVideoI18n() {
setText("wfm-video-edit-duplicate-btn", "videoEditDuplicate");
setText("wfm-video-edit-delete-btn", "videoEditDelete");
setText("wfm-video-edit-clear-btn", "videoEditClearBtn");
setTitle("wfm-video-edit-undo-btn", "videoEditUndo");
setTitle("wfm-video-edit-redo-btn", "videoEditRedo");
setText("wfm-video-edit-audio-title", "videoEditAudioTitle");
setText("wfm-video-edit-keep-audio-label", "videoEditKeepAudio");
setText("wfm-video-edit-orig-vol-label", "videoEditOriginalVolume");
setText("wfm-video-edit-bgm-btn", "videoEditBgmSelect");
setTitle("wfm-video-edit-bgm-clear", "videoEditBgmRemove");
setText("wfm-video-edit-bgm-vol-label", "videoEditBgmVolume");
setText("wfm-video-edit-bgm-offset-label", "videoEditBgmOffset");
setText("wfm-video-edit-bgm-hint", "videoEditBgmHint");
setText("wfm-video-source-add-to-edit", "videoSourceAddToEdit");
setText("wfm-video-source-label", "videoSourceLabel");
setText("wfm-video-source-hint", "videoSourceHint");
+47 -11
View File
@@ -1661,7 +1661,7 @@
<h3 id="wfm-help-settings-card-general-title">General</h3>
<ul>
<li id="wfm-help-settings-1">UI language and summary language</li>
<li id="wfm-help-settings-2">Theme selection (13 presets) with customization: color overrides, background patterns (stripes, polka dot, checkerboard, SVG tiling), and font selection (16 Google Fonts including Japanese display fonts)</li>
<li id="wfm-help-settings-2">Theme selection (13 presets) with customization: color overrides, background patterns (stripes, polka dot, checkerboard, SVG tiling), and font selection (15 Google Fonts including Japanese display fonts, plus the system default)</li>
<li id="wfm-help-settings-9">All sections are collapsible accordion panels</li>
</ul>
</div>
@@ -1671,20 +1671,25 @@
<li id="wfm-help-settings-4">ComfyUI connection URL and test</li>
<li id="wfm-help-settings-5">AI Assistant settings have moved out of this tab &mdash; the Prompt tab, Tagger tab, and AI TOOL tab each have their own &#9881; Settings button (backend, URL, model, Thinking mode, Max tokens, unload model), saved independently</li>
<li id="wfm-help-settings-7">Eagle integration URL and auto-save toggle</li>
<li id="wfm-help-settings-22">G'MIC-Qt Integration: path to the G'MIC-Qt executable (gmic_qt.exe) used by the Image Edit tab's G'MIC filter (Edit with G'MIC); Save stores it — see the Image Edit help for installing G'MIC-Qt</li>
<li id="wfm-help-settings-17">Decision Model: its own section, independent of the AI TOOL / Tagger LLM/VLM backends so it can be used alongside them — backend, URL, model and a default probability threshold (answers below it are treated as suggestions to confirm). Unsloth: Laya models (laya / laya-multilingual / laya-english / laya-typed-decisions); requires Settings → API → Decision API → Serve requests in Unsloth Desktop (otherwise HTTP 404 "The Decision API is off") and either UNSLOTH_API_KEY in .env or Keyless API access → "Chat and inference". Ollama (0.35+): lists installed models whose capabilities include "decision" (tev1, tev1:0.8b, nimble — ollama pull first; ↻ refreshes the list); no key needed. Test sends one yes/no question with the unsaved form values; the first request loads the model.</li>
</ul>
</div>
<div class="wfm-help-card">
<h3 id="wfm-help-settings-card-workflow-title">Workflow Defaults</h3>
<ul>
<li id="wfm-help-settings-3">Workflow data folder path</li>
<li id="wfm-help-settings-19">Gallery Output Folder: the ComfyUI output folder the Gallery tab browses — leave blank to use ComfyUI's default output directory; Apply takes effect immediately, Default resets it, and the folder currently in use is shown below the field</li>
<li id="wfm-help-settings-6">Default workflow setting</li>
<li id="wfm-help-settings-10">Wildcard Integration: link WFS wildcard dir to ComfyUI-Impact-Pack wildcards (junction/symlink); existing files migrated automatically</li>
<li id="wfm-help-settings-20">Default Checkpoint: when "Force default checkpoint on workflow load" is on, every checkpoint loader node's model is replaced with the chosen checkpoint whenever a workflow is loaded into GenerateUI (a toast shows when it's applied) — avoids errors or garbled images caused by a model name saved in the workflow that you don't have; Refresh reloads the checkpoint list, Save stores the setting</li>
<li id="wfm-help-settings-10">Wildcard Integration: link WFS wildcard dir to ComfyUI-Impact-Pack wildcards (directory junction on Windows / symlink on other OS); existing files migrated automatically</li>
</ul>
</div>
<div class="wfm-help-card">
<h3 id="wfm-help-settings-card-editor-title">Editor Customization</h3>
<ul>
<li id="wfm-help-settings-11">Text Size: one slider (10–28 px) adjusts font size for all prompt and chat textareas at once — Generate UI positive/negative prompts, AI Assistant chat input, Preset positive/negative, Wildcard prompt and file editor, and Metadata prompt full preview; takes effect immediately and saved with Save Settings</li>
<li id="wfm-help-settings-18">Video Volume: one slider (0–100%) sets the playback volume of the Video tab preview and the Gallery detail / lightbox / compare video players; the level set here (or changed with any video's own controls) carries over to the next video opened anywhere in the app</li>
<li id="wfm-help-settings-12">RAW JSON Colors: customize the 6 syntax highlight colors for the Raw JSON editor in Generate UI — Default Text, Name/Scheduler, Title, Width/Height, Prompt/Text, Image/File; changes apply immediately on color pick; Reset Defaults restores the original color scheme</li>
<li id="wfm-help-settings-14">GenerateUI Model Tab Highlight: pick the color used to mark Checkpoint/VAE/Diffusion Model/Text Encoder/ControlNet/Hypernetwork labels and the LoRA column header in GenerateUI's Model tab when the loaded workflow actually contains that type of node; default bright green; changes apply immediately, Reset Defaults restores it</li>
</ul>
@@ -1692,10 +1697,11 @@
<div class="wfm-help-card">
<h3 id="wfm-help-settings-card-misc-title">Data &amp; Misc</h3>
<ul>
<li id="wfm-help-settings-8">Data Management: export core settings/metadata/prompts to JSON / import to restore</li>
<li id="wfm-help-settings-15">Full Backup (ZIP): unlike the JSON export above, also includes folder-based data (AI Skills, Lab Plans, Style presets) and the Tagger database; Restore overwrites existing files — restart ComfyUI afterward</li>
<li id="wfm-help-settings-16">Two opt-in checkboxes above the Full Backup buttons additionally include ComfyUI's default workflows folder and the Impact Pack's wildcard folder in the export — both export-only, never restored automatically; copy the ZIP's _external/ folder back by hand if you want them</li>
<li id="wfm-help-settings-8">Data Management: export core settings, metadata, and prompts to a JSON file, or import to restore. After importing, restart ComfyUI.</li>
<li id="wfm-help-settings-15">Full Backup (ZIP): unlike the JSON export above, also includes folder-based data (AI Skills, Lab Plans, Video Plans, Video Edit projects, Style presets) and the Tagger database — everything the JSON export can't represent. Restoring overwrites existing files; restart ComfyUI afterward.</li>
<li id="wfm-help-settings-16">Two opt-in checkboxes above the Full Backup buttons additionally include ComfyUI's default workflows folder and the Impact Pack's wildcard folder in the export — both export-only, never restored automatically; copy the ZIP's _external/ folder back by hand if you want them. A third checkbox includes the clips and BGM referenced by saved Video Edit projects (projects only store references to files in ComfyUI's input folder) — unlike the other two, these are restored into the input folder, skipping any file that already exists there.</li>
<li id="wfm-help-settings-13">CivitAI Host: choose between civitai.com (SFW only) and civitai.red (unrestricted) — determines which site opens when clicking a model link in the Models tab CivitAI panel</li>
<li id="wfm-help-settings-21">CivitAI API Key: optional key for authenticated CivitAI access from the Models tab (get it at civitai.com → Account Settings → API Keys); stored in settings.json but left out of the JSON export and Full Backup, and the CIVITAI_API_KEY environment variable takes priority when set</li>
</ul>
</div>
</div>
@@ -1928,7 +1934,7 @@
</div>
<div class="wfm-help-card">
<h3 id="wfm-help-ai-card-settings-title">Settings pane</h3>
<p id="wfm-help-ai-6">Settings pane: choose backend (Ollama / LM Studio / Lemonade / Unsloth), set API URL, test connection, select model, and use the Unload Model button (next to Test connection) to free the loaded model from VRAM/memory &mdash; Ollama uses keep_alive:0, LM Studio uses /api/v1/models/unload, Lemonade uses /api/v1/unload; Unsloth has no unload API, so this shows an unsupported message for it; Unsloth always requires an API key even for local access — set UNSLOTH_API_KEY in a .env file in the plugin folder (copy .env.example) and restart ComfyUI, the key is read server-side and never sent to the browser; Unsloth also requires "Auto model switching (OpenAI API)" to be turned ON in the Unsloth Desktop app itself (its own Settings screen) — without it, requests fail with HTTP 404 unless a model was already manually loaded in the Unsloth Studio UI; Unsloth reasoning models return their thinking in a separate reasoning_content field rather than inline &lt;think&gt; tags, so a low Max tokens can make the model spend its whole budget reasoning and return an empty reply — increase Max tokens (2048+) if replies come back blank; Generation section — Thinking mode toggle (off by default; Ollama receives a `think` flag, and regardless of backend any &lt;think&gt;...&lt;/think&gt; block is stripped from the output when off, so reasoning models don't clutter Translation/Chat/TOOLS results) and Max tokens (Ollama: num_predict, LM Studio/Lemonade/Unsloth: max_tokens; 0 = unlimited); configure Free language names; Chat Image Generation section lets you enable a dedicated workflow for Chat-triggered text-to-image generation instead of the one currently loaded in GenerateUI; Chat I2I Generation section is the same but for image-to-image generation when an image is attached; settings are shared with the Library panel's A tab via localStorage</p>
<p id="wfm-help-ai-6">Settings pane: choose backend (Ollama / LM Studio / Lemonade / Unsloth), set API URL, test connection, select model, and use the Unload Model button (next to Test connection) to free the loaded model from VRAM/memory &mdash; Ollama uses keep_alive:0, LM Studio uses /api/v1/models/unload, Lemonade uses /api/v1/unload; Unsloth has no unload API, so this shows an unsupported message for it; Unsloth needs either an API key or Keyless API access — set UNSLOTH_API_KEY in a .env file in the plugin folder (copy .env.example) and restart ComfyUI (the key is read server-side and never sent to the browser), or leave it unset and turn on Keyless API access → "Chat and inference" in Unsloth Desktop's Settings → API (requests are then sent without a key; note that setting also opens the API to your private LAN, and keep "Allow tools" off); Unsloth also requires "Auto model switching (OpenAI API)" to be turned ON in the Unsloth Desktop app itself (its own Settings screen) — without it, requests fail with HTTP 404 unless a model was already manually loaded in the Unsloth Studio UI; Unsloth reasoning models return their thinking in a separate reasoning_content field rather than inline &lt;think&gt; tags, so a low Max tokens can make the model spend its whole budget reasoning and return an empty reply — increase Max tokens (2048+) if replies come back blank; Generation section — Thinking mode toggle (off by default; Ollama receives a `think` flag, and regardless of backend any &lt;think&gt;...&lt;/think&gt; block is stripped from the output when off, so reasoning models don't clutter Translation/Chat/TOOLS results) and Max tokens (Ollama: num_predict, LM Studio/Lemonade/Unsloth: max_tokens; 0 = unlimited); configure Free language names; Chat Image Generation section lets you enable a dedicated workflow for Chat-triggered text-to-image generation instead of the one currently loaded in GenerateUI; Chat I2I Generation section is the same but for image-to-image generation when an image is attached; settings are shared with the Library panel's A tab via localStorage</p>
</div>
</div>
@@ -2105,7 +2111,7 @@
<div class="wfm-help-page" id="wfm-help-page-video">
<h2 id="wfm-help-video-title">Video Tab</h2>
<p id="wfm-help-video-desc">A dedicated video generation tab for MiniMax H3 (Image to Video), LTX-2.5 (Text / Image / First & Last Frame to Video), and Wan2.2 (Text to Video) workflows, independent of the GenerateUI tab. The center panel's Plan / Edit / Asset subtabs handle batch generation, clip editing, and browsing generated videos; the sidebar's Project subtab handles saved plans.</p>
<p id="wfm-help-video-desc">A dedicated video generation tab for MiniMax H3 (Image to Video), LTX-2.5 (Text / Image / First &amp; Last Frame to Video), and Wan2.2 (Text to Video) workflows, independent of the GenerateUI tab. The center panel's Plan / Edit / Asset subtabs handle batch generation, clip editing, and browsing generated videos; the sidebar's Project subtab handles saved plans (while Plan is active) or saved Edit projects (while Edit is active).</p>
<div class="wfm-help-card">
<ul>
<li id="wfm-help-video-1">Plan subtab: the topbar's settings row holds Aspect Ratio / Megapixels / Multiple / Duration (seconds) / Noise Seed as this plan's base settings — Duration always reflects the currently selected block (each block keeps its own value), the rest are plan-wide. A second row below adds Width / Height / Turbo Mode / Strength / Steps / Prompt Enhance, each enabled only for workflows that actually support it (see helpVideo2). Split the timeline into blocks with the + Split / + Block / Delete buttons — each block has its own Prompt, Duration, and First/Last Image — First Image can be None, an explicit image, or "use previous block's last frame" (an explicit Last Image on the previous block always wins over extracting one from its generated video). Click Run to generate every block in order; Pause/Cancel work mid-run.</li>
@@ -2116,9 +2122,9 @@
<li id="wfm-help-video-6">Frame tab (right panel): play/seek the preview video to the position you want, click Capture Frame to grab that frame client-side, then Save to Output to write it as a PNG into ComfyUI's output folder — check "Add to Video Assets group" first to also tag the saved frame into Gallery's Video Assets group.</li>
<li id="wfm-help-video-7">GIF tab (right panel): set Start/End (seconds), FPS, and an optional Max Width, then Convert to GIF — the server decodes the video and encodes an animated GIF into ComfyUI's output folder.</li>
<li id="wfm-help-video-8">Chaining: setting a block's First Image to "use previous block's last frame" continues smoothly from the previous clip — if that previous block has its own explicit Last Image, that image is used directly; otherwise the actual last frame of its generated video is extracted automatically.</li>
<li id="wfm-help-video-9">Project tab (sidebar): browse every saved plan file (thumbnail, block count, last updated), click one to load it into the Plan editor, or delete it with the × button; clicking the same tab again collapses the panel.</li>
<li id="wfm-help-video-9">Project tab (sidebar): lists saved Video Plans while the center Plan subtab is active, or saved Edit projects while the center Edit subtab is active (thumbnail/block or clip count, last updated) — click one to load it, or delete it with the × button; clicking the same tab again collapses the panel.</li>
<li id="wfm-help-video-10">Highlight color: the Prompt / Duration / First Image / Last Image / Aspect Ratio / Megapixels / Multiple / Noise Seed / Width / Height / Turbo Mode / Strength / Steps / Prompt Enhance labels turn the same highlight color as the GenerateUI Model tab (customizable in the Settings tab) when the loaded workflow actually supports that field — e.g. for a Wan2.2 workflow (no ResolutionSelector node), the Aspect Ratio / Megapixels / Multiple labels stay unhighlighted while Width / Height light up instead.</li>
<li id="wfm-help-video-11">Edit subtab: a lightweight clip timeline for trimming and concatenating video/image clips outside of generation. Add clips via the Video Source panel's "Add to Edit" button or the Asset subtab's "Send to Edit" button, then drag blocks to reorder (or use the ◀/▶ toolbar buttons), Duplicate/Delete the selected clip, and Clear to empty the timeline. The trim panel for the selected video clip shows a time ruler, a draggable playhead synced to the Source preview, and draggable in/out handles with time badges — image clips instead get a simple hold-duration field. ▶ Preview plays every ready clip back-to-back (trimmed to its in/out points) in the right-hand result preview; Export builds and runs the actual ComfyUI graph (trim → concatenate if more than one clip → save), tagging the result into the Video Temp group automatically. Exporting more than one clip currently drops audio, since ComfyUI's concatenation node requires uniform audio across all inputs.</li>
<li id="wfm-help-video-11">Edit subtab: a lightweight clip timeline for trimming and concatenating video/image clips outside of generation. Add clips via the Video Source panel's "Add to Edit" button or the Asset subtab's "Send to Edit" button, then drag blocks to reorder (or use the ◀/▶ toolbar buttons), Duplicate/Delete the selected clip, and Clear to empty the timeline (this also resets the Audio section). The trim panel for the selected video clip shows a time ruler, a draggable playhead synced to the Source preview, and draggable in/out handles with time badges — image clips instead get a simple hold-duration field. ▶ Preview plays every ready clip back-to-back (trimmed to its in/out points) in the right-hand result preview; Export builds and runs the actual ComfyUI graph (trim → concatenate if more than one clip → save), tagging the result into the Video Temp group automatically. The Audio section of the export panel controls the soundtrack: "Keep original audio" (with its own volume in dB) keeps each clip's audio aligned to its segment even across multiple clips (image clips contribute silence), and "BGM..." lays one audio file over the whole video with its own volume and start offset (a longer BGM is cut to the video length). Each clip can also carry text overlays ("+ Add text" under the trim panel: text, start/end seconds within the clip, size as % of frame height, color, 9-position anchor, outline, background box); they are previewed live over the Source/Result panes and burned in on Export as a final pass. The BGM can be chosen with "BGM...", by dropping an audio/video file onto the Audio section, or with "Set as BGM" on a video in the Asset subtab (its audio track is used). Video clips can also be cropped: "Edit crop" shows a draggable/resizable rectangle over the Source preview (optionally locked to 16:9 / 9:16 / 1:1 / 4:3 / 3:4) — "Done" keeps it, "Cancel" restores the crop from before editing, and switching to another subtab/tab ends editing like Done; the export size becomes the first video clip's cropped size, and other clips of a different size are center-fit to it. ↶/↷ in the timeline toolbar (or Ctrl+Z / Ctrl+Y) undo and redo edits to clips, trims, crops, text and audio settings. The export panel's Save/Save As/Load buttons persist the timeline itself (clip order, trim points, crop, text overlays, audio/BGM settings, and which uploaded file each clip and the BGM refer to) as a JSON project file on the server. The project only stores references to those files in ComfyUI's input folder, so Load re-fetches each clip from its saved server reference and re-probes it — the original files must still exist (Settings → Full Backup can bundle them with "Include media used by Video Edit projects"). These saved projects also show up in the sidebar's Project tab whenever the Edit subtab is active (see the Project tab item).</li>
<li id="wfm-help-video-12">Asset subtab (center panel, next to Plan/Edit): browses every video/image tagged into Video Assets or Video Temp (via the group dropdown, or "All Video Assets" for both). Search filters by filename; the ▦/☰ buttons switch between a thumbnail grid and a table (filename/updated/tags). Selecting an item shows its tags/memo, loads it into the center preview, and offers "Send to Edit" (queues it as a clip in the Edit subtab) plus, for images only, "Set as First Image"/"Set as Last Image" (uploads it straight into the currently-selected Plan block, switching to the Plan subtab).</li>
</ul>
</div>
@@ -2942,6 +2948,8 @@
<button type="button" id="wfm-video-edit-duplicate-btn" class="wfm-btn wfm-btn-xs">Duplicate</button>
<button type="button" id="wfm-video-edit-delete-btn" class="wfm-btn wfm-btn-xs">Delete</button>
<button type="button" id="wfm-video-edit-preview-btn" class="wfm-btn wfm-btn-xs">▶ Preview</button>
<button type="button" id="wfm-video-edit-undo-btn" class="wfm-btn wfm-btn-xs" title="Undo (Ctrl+Z)" disabled>↶</button>
<button type="button" id="wfm-video-edit-redo-btn" class="wfm-btn wfm-btn-xs" title="Redo (Ctrl+Y)" disabled>↷</button>
<span class="wfm-i2i-status wfm-video-edit-total-duration" id="wfm-video-edit-total-duration" style="margin-left:auto;"></span>
<button type="button" id="wfm-video-edit-clear-btn" class="wfm-btn wfm-btn-xs wfm-btn-danger">Clear</button>
</div>
@@ -2955,9 +2963,37 @@
<div style="display:flex;gap:6px;">
<button type="button" id="wfm-video-edit-save-btn" class="wfm-btn wfm-btn-sm" style="flex:1;">Save</button>
<button type="button" id="wfm-video-edit-saveas-btn" class="wfm-btn wfm-btn-sm" style="flex:1;">Save As</button>
<button type="button" id="wfm-video-edit-load-btn" class="wfm-btn wfm-btn-sm" style="flex:1;">Load</button>
<!-- File pickers are native <label>-wrapped inputs (same as the Video Source drop zone),
not a button calling input.click() from JS — that programmatic path can be refused
by the browser without any error (confirmed on a user's machine for the BGM picker). -->
<label id="wfm-video-edit-load-btn" class="wfm-btn wfm-btn-sm wfm-video-edit-file-btn" style="flex:1;" tabindex="0" role="button">Load<input type="file" id="wfm-video-edit-load-file-input" accept="application/json,.json" style="display:none;"></label>
</div>
<div class="wfm-video-edit-audio-panel">
<div class="wfm-video-edit-section-head"><span id="wfm-video-edit-audio-title">Audio</span></div>
<label class="wfm-video-checkbox-inline">
<input type="checkbox" id="wfm-video-edit-keep-audio" checked> <span id="wfm-video-edit-keep-audio-label">Keep original audio</span>
</label>
<div class="wfm-video-edit-audio-row">
<label for="wfm-video-edit-orig-vol" id="wfm-video-edit-orig-vol-label">Original volume</label>
<input type="range" id="wfm-video-edit-orig-vol" min="-30" max="12" step="1" value="0">
<span class="wfm-video-edit-audio-val" id="wfm-video-edit-orig-vol-val">0 dB</span>
</div>
<div class="wfm-video-edit-bgm-file">
<label class="wfm-btn wfm-btn-xs wfm-video-edit-file-btn" tabindex="0" role="button"><span id="wfm-video-edit-bgm-btn">BGM...</span><input type="file" id="wfm-video-edit-bgm-input" accept="audio/*,video/mp4,.mp3,.wav,.m4a,.aac,.flac,.ogg,.opus,.mp4" style="display:none;"></label>
<span class="wfm-i2i-status wfm-video-edit-bgm-name" id="wfm-video-edit-bgm-name"></span>
<button type="button" id="wfm-video-edit-bgm-clear" class="wfm-btn wfm-btn-xs" title="Remove BGM">✕</button>
</div>
<div class="wfm-i2i-status wfm-video-edit-bgm-hint" id="wfm-video-edit-bgm-hint">Drop an audio file here, or use &quot;Set as BGM&quot; on a video in Asset</div>
<div class="wfm-video-edit-audio-row">
<label for="wfm-video-edit-bgm-vol" id="wfm-video-edit-bgm-vol-label">BGM volume</label>
<input type="range" id="wfm-video-edit-bgm-vol" min="-30" max="12" step="1" value="-6">
<span class="wfm-video-edit-audio-val" id="wfm-video-edit-bgm-vol-val">-6 dB</span>
</div>
<div class="wfm-video-edit-audio-row">
<label for="wfm-video-edit-bgm-offset" id="wfm-video-edit-bgm-offset-label">BGM start (s)</label>
<input type="number" id="wfm-video-edit-bgm-offset" class="wfm-input" min="0" step="0.1" value="0">
</div>
</div>
<input type="file" id="wfm-video-edit-load-file-input" accept="application/json,.json" style="display:none;">
<button type="button" id="wfm-video-edit-export-btn" class="wfm-btn wfm-btn-primary" style="width:100%;margin-top:8px;">Export</button>
<div class="wfm-gen-progress-container" style="margin-top:8px;">
<div class="wfm-gen-progress-track">
+20 -13
View File
@@ -3253,17 +3253,16 @@ function _extractPromptsFromNodeSet(nodes, links) {
continue;
}
if (!_isTextEncoderNode(type)) continue;
const text = n.widgets_values?.[0];
if (text && typeof text === "string") { textMap.set(n.id, text); }
else if (Array.isArray(n.inputs)) {
const textInput = n.inputs.find(inp => inp.name === "text" || inp.name === "text_g" || inp.name === "prompt");
if (textInput?.link != null) {
const originId = linkOrigin.get(textInput.link);
const originSlot = linkSlot.get(textInput.link) ?? 0;
const text2 = _resolveLinkedTextInNodeSet(nodeMap, linkOrigin, linkSlot, originId, originSlot);
if (text2) textMap.set(n.id, text2);
}
}
// リンク化されたtextウィジェットはwidgets_values[0]に古い入力値が残る(実行時には使われない)ため、
// リンク先の解決結果を優先する
const textInput = Array.isArray(n.inputs)
? n.inputs.find(inp => inp.name === "text" || inp.name === "text_g" || inp.name === "prompt")
: null;
const linkedText = textInput?.link != null && linkOrigin.has(textInput.link)
? _resolveLinkedTextInNodeSet(nodeMap, linkOrigin, linkSlot, linkOrigin.get(textInput.link), linkSlot.get(textInput.link) ?? 0)
: null;
const text = linkedText || n.widgets_values?.[0];
if (text && typeof text === "string") textMap.set(n.id, text);
}
const pos = new Set(), neg = new Set();
let foundSampler = false;
@@ -3273,11 +3272,13 @@ function _extractPromptsFromNodeSet(nodes, links) {
// SamplerCustomAdvanced doesn't hold positive/negative directly — resolve through its
// "guider" input (CFGGuider/DualCFGGuider/BasicGuider) instead.
let inputsToScan = n.inputs;
let viaBasicGuider = false;
const guiderInput = n.inputs.find(inp => inp.name === "guider");
if (guiderInput?.link != null) {
const guiderId = linkOrigin.get(guiderInput.link);
const guiderNode = guiderId != null ? nodeMap.get(guiderId) : null;
if (Array.isArray(guiderNode?.inputs)) inputsToScan = guiderNode.inputs;
viaBasicGuider = guiderNode?.type === "BasicGuider";
}
for (const inp of inputsToScan) {
if (!inp || inp.link == null) continue;
@@ -3285,7 +3286,9 @@ function _extractPromptsFromNodeSet(nodes, links) {
if (originId == null) continue;
const name = inp.name ?? "";
// DualCFGGuider (HiDream E1): cond1 carries the positive-derived conditioning.
const isPos = name === "positive" || name === "cond1" || name.startsWith("positive");
// BasicGuider (Flux/Ming Image etc., no negative): its single "conditioning" input is positive.
const isPos = name === "positive" || name === "cond1" || name.startsWith("positive")
|| (viaBasicGuider && name === "conditioning");
const isNeg = name === "negative" || name.startsWith("negative");
if (!isPos && !isNeg) continue;
// TextEncodeMageFlowEdit / TextEncodeBooguEdit — 1ノードでprompt/negative_prompt両方を
@@ -3428,14 +3431,18 @@ function _extractPromptsAPI(wf) {
// Guider node (CFGGuider/DualCFGGuider/BasicGuider) via its "guider" input. Scan that
// node's inputs instead so the loop below can find the positive/negative-role keys.
let inputsToScan = n.inputs ?? {};
let viaBasicGuider = false;
if (Array.isArray(n.inputs?.guider)) {
const guiderNode = wf[String(n.inputs.guider[0])];
if (guiderNode?.inputs) inputsToScan = guiderNode.inputs;
viaBasicGuider = guiderNode?.class_type === "BasicGuider";
}
for (const [key, val] of Object.entries(inputsToScan)) {
if (!Array.isArray(val)) continue;
// DualCFGGuider (HiDream E1): cond1 carries the positive-derived conditioning.
const isPos = key === "positive" || key === "cond1" || key.startsWith("positive");
// BasicGuider (Flux/Ming Image etc., no negative): its single "conditioning" input is positive.
const isPos = key === "positive" || key === "cond1" || key.startsWith("positive")
|| (viaBasicGuider && key === "conditioning");
const isNeg = key === "negative" || key.startsWith("negative");
if (!isPos && !isNeg) continue;
// TextEncodeMageFlowEdit / TextEncodeBooguEdit — 1ノードでprompt/negative_prompt両方を