feat(audio): GPT-SoVITS setup card and character import

Uses Anomalous_TTS interface v3 so a new user can get going without
editing extra_model_paths.yaml or sorting model folders by hand.

- Setup card on the GPT-SoVITS studio: character libraries (add/remove),
  pretrained files (per-item or all download with progress, or use an
  existing GPT-SoVITS package) and missing Python packages with the
  install command. Collapses to one line when ready.
- Import form: drop files, pick them in the browser, or pick them on the
  ComfyUI machine (server-side path picker, no upload). Name, SoVITS
  version, clip length, lines, main voice and language are filled in from
  the node's inspect; emotions per clip. Uploads go in 8 MB chunks.
- "Add files" on a card, or a drop onto it, adds clips and emotions to
  that character without replacing its main voice.
- Hidden for nodes without v3; disabled when the page is not opened on
  the ComfyUI machine.
- New modules: tts_setup_api.js (calls + pure form rules),
  ui_tts_setup.js, ui_tts_import.js, ui_tts_path_picker.js and
  styles/12-tts-setup.css. ui_audio_studio.js only wires them in.

Tested: node tests/tts_setup_api.mjs (new, local) plus the audio engine,
node target and script rule tests. Isolated ComfyUI with both plugins in
the browser: setup card for a fresh user, package as pretrained source,
library add, import from local paths (lines found in all.txt), file drop
onto a card (chunked upload, add mode), cancel discards the upload.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

ARCHITECTURE.md

@@ -134,10 +134,12 @@ DOM or live LiteGraph state.
   voice-group shape (`node_value` = what the node's voice widget takes), so cards,
   sidebar and Script Director share one implementation. A missing engine stays in
   the switch marked "not installed" with install steps; nothing else depends on it.
-  GPT-SoVITS data comes only from the node's HTTP contract, version 2
-  (`/anomalous_tts/characters`, `/audio`, `/settings`; the node repo's
-  `docs/INTERFACE.md`, mirrored as the project doc `anomalous-tts-interface.md`);
-  Anomalous never reads or writes its model folders. The character list is a
+  GPT-SoVITS data comes only from the node's HTTP contract, version 3
+  (`/anomalous_tts/characters`, `/audio`, `/settings`, `/status`, the library,
+  pretrained, browse and import routes; the node repo's `docs/INTERFACE.md`,
+  mirrored as the project doc `anomalous-tts-interface.md`); Anomalous never
+  reads or writes its model folders. `loadGptSovitsStatus` caches the setup status
+  with the engine data (null for a node without v3, which hides setup and import). The character list is a
   summary without file lists; `fetchGptSovitsCharacter` gets one character's files.
   Engine presence and loaded voice groups are cached (`MAX_AGE_MS`) so re-renders
   from sidebar clicks, domain switches and the studio + sidebar pair cost no
@@ -147,6 +149,16 @@ DOM or live LiteGraph state.
   `ui_audio_tts_editor.js` edits a character's emotion references through that API:
   it opens at once, loads the character's audio list in the background (save waits
   for it), and keeps settings fields it does not know.
+  `tts_setup_api.js` holds the other v3 calls (libraries, pretrained sources and
+  downloads, folder browse, chunked upload, inspect, commit, discard) and the pure
+  import-form rules (`importKind`, `buildImportBody`, `importProblem`,
+  `setupSummary`); no DOM. `ui_tts_setup.js` is the setup card at the top of the
+  GPT-SoVITS studio (polls the status only while a download runs and it is in the
+  page). `ui_tts_import.js` is the import form (new character or `target` = add
+  files; rows are built once and updated in place; closing discards unfinished
+  uploads) and `bindTtsFileDrop`, which turns OS file drops on the studio or a card
+  into that form. `ui_tts_path_picker.js` picks server-side folders or files
+  through the node's browse route, since the browser cannot see local paths.
 - `audio_node_targets.js` is the single table of canvas nodes the audio studio
   writes into and what each takes: F5-TTS (`F5TTSAudio`, `F5TTSAudioAdvanced`,
   `F5TTSAudioFromModel`) takes a whole character (main voice in `sample`,

CHANGELOG.md

@@ -10,6 +10,7 @@
 - **Clear drag rules (拖动规则明确)**: drag a character card's header onto an F5-TTS node (sets the main voice), or one row onto a Load Audio node (uses that exact clip). The hint says what releasing will do; unsupported nodes are refused with the reason instead of being guessed. A single emotion clip is never written into F5-TTS's `sample`, which would break `{emotion}` tags.
 - **Usage guide (使用说明)**: in the audio page, the header **!** opens a five-step how-to; the full guide is `docs/guides/audio-studio.md`.
 - **Engine switch: F5-TTS / GPT-SoVITS (引擎切换)**: the audio page now manages voices per engine. Installed engines are detected at runtime; a missing one stays visible as "not installed" with install steps instead of disappearing, and nothing else depends on it. GPT-SoVITS characters come from the separate Anomalous_TTS node pack: preview reference clips, pick a reference clip per emotion ("Edit emotions", saved through the node's API), drag a character onto the node, and push scripts with the Script Director.
+- **GPT-SoVITS setup and import (GPT-SoVITS 准备与导入)**: a setup card shows the character libraries, pretrained files and missing Python packages; add a library, download pretrained files with progress, or use an existing GPT-SoVITS package, all without editing `extra_model_paths.yaml` or restarting. **Import character** builds a character from dropped or picked GPT / SoVITS weights, reference clips and annotation files: the name, SoVITS version, clip lengths and lines are filled in automatically, and everything is copied (originals stay). Drop files on a card, or use **Add files**, to add clips and emotions to an existing character. Needs Anomalous_TTS with interface version 3; file actions only work on the computer running ComfyUI.
 
 ### Fixes
 - All-in-one loader recipes and cross-node prompt injection use one verified adapter table; prompt text never crosses between positive and negative.

docs/guides/audio-studio.md

@@ -110,10 +110,33 @@ Load Audio 正好相反：它只用一条具体的音频，所以要拖某一行
 
 ## 6. GPT-SoVITS（Anomalous TTS）
 
-### 角色放在哪
+### 准备状态
+
+切到 GPT-SoVITS 后，最上面是 **GPT-SoVITS 准备状态**。全部就绪时它收成一行，点一下展开：
+
+- **角色库**：节点从哪些文件夹里找角色，每个有几个角色。**+ 添加角色库** 选一个你已经整理好的文件夹（里面每个子文件夹是一个角色），马上生效，不用重启，也不用改 `extra_model_paths.yaml`。在这里添加的可以再移除；移除只是不再列出，不会删文件。
+- **底模**：每个底模是已就绪、缺少还是下载中。可以单个 **下载**、**全部下载**，或者 **使用已有的 GPT-SoVITS 整合包**（选整合包里的 `GPT_SoVITS` 文件夹），就不用再下一遍。只说日语的角色不需要中文底模；标“可选”的缺了也能用。
+- **Python 包**：缺了哪种语言要的包，会给出安装命令，复制到命令行运行后重启 ComfyUI。
+
+这些操作会读写运行 ComfyUI 那台电脑上的文件，所以只能在那台电脑上用；从别的设备打开时按钮是灰的。
+
+### 导入角色（新用户从这里开始）
+
+点 **+ 导入角色**，或者直接把文件拖到音频页上：
+
+1. 放进 GPT 权重（`.ckpt`）、SoVITS 权重（`.pth`）、参考音频，最好再加上 GPT-SoVITS 的标注文件（`.list` 或同格式的 `.txt`）。可以拖进来、用 **选择文件…**，或者用 **从本机路径选择…** 直接选电脑上的文件（比如整合包里的 `GPT_weights_v2`，不用上传，更快）。
+2. 表单会自动填好：角色名（取自 GPT 文件名）、SoVITS 的版本（v3 / v4 暂不支持，会提示）、每段音频的长度和台词（从标注文件里找）、主声音（第一段 3~10 秒且有台词的）、语言。都可以改。
+3. 其他音频填上情绪名（例如“开心”），就成了 `{开心}`。
+4. 选放到哪个角色库，点 **导入**。
+
+节点会建好 `角色名/GPT_weights/`、`SoVITS_weights/`、`参考音频/` 和设置文件。**一律复制，原文件不动**；中途出错时角色库保持原样，不会留下半个角色。
+
+给已有角色加音频或情绪：点卡片上的 **添加文件**，或者把文件拖到那张卡片上。已有的文件不会被覆盖，主声音也不会被换掉（除非你在表单里点了“主声音”）。
+
+### 角色文件夹的规则
 
 - 一个角色就是一个文件夹，里面至少有一个 `.ckpt`（GPT 权重）和一个 `.pth`（SoVITS 权重）。
-- 文件夹放在 `ComfyUI/models/gpt_sovits/` 下，或者在 `extra_model_paths.yaml` 里加一个 `gpt_sovits` 目录，直接用你原来的模型文件夹，不用复制。
+- 文件夹放在任何一个角色库里：默认的 `ComfyUI/models/gpt_sovits/`、在上面添加的角色库，或者 `extra_model_paths.yaml` 里的 `gpt_sovits` 目录。
 - 同一个角色有日配、中配两个版本时，会显示成 `阿罗娜/日配`、`阿罗娜/中配` 两个角色。
 
 ### 编辑情绪
@@ -150,3 +173,5 @@ Load Audio 正好相反：它只用一条具体的音频，所以要拖某一行
 | 提示“这是 GPT-SoVITS 的声音，F5-TTS 节点用不了” | 两种引擎的声音不能混用。切到对应的引擎，拖到它自己的节点上 |
 | GPT-SoVITS 角色卡片上提示“没有找到合适的主声音” | 角色文件夹里没有 3~10 秒、找得到台词的音频。点“编辑情绪”手动选一条 |
 | 新放进去的 GPT-SoVITS 角色或音频没出现 | 点 **刷新音色**。节点那边在 ComfyUI 里按 **R** 刷新下拉框 |
+| 看不到“准备状态”和“导入角色” | Anomalous_TTS 版本太旧（没有接口版本 3）。更新 Anomalous_TTS 后重启 ComfyUI |
+| “导入角色”是灰的 | 页面是从别的设备打开的。导入会写文件，只能在运行 ComfyUI 的电脑上做 |

web/modules/audio_engines.js

@@ -174,9 +174,29 @@ async function loadGroups(engineId) {
 export function invalidateEngineCache({ rescan = false } = {}) {
     loadCache.clear();
     installCache = null;
+    statusCache = null;
     if (rescan) rescanNext = true;
 }
 
+let statusCache = null; // { at, promise } — GPT-SoVITS setup status
+
+/**
+ * The GPT-SoVITS node's setup status (libraries, pretrained files, packages; interface v3),
+ * or null for a node that predates it (no `/anomalous_tts/status`). Cached like loadEngine;
+ * `force` skips the cache (download progress polling).
+ */
+export function loadGptSovitsStatus({ force = false } = {}) {
+    if (!force && fresh(statusCache)) return statusCache.promise;
+    const promise = getJson('/anomalous_tts/status').then(({ ok, status, data }) => {
+        if (status === 404) return null;
+        if (!ok || !data) throw new Error(`HTTP ${status}`);
+        return data;
+    });
+    statusCache = { at: Date.now(), promise };
+    promise.catch(() => { if (statusCache?.promise === promise) statusCache = null; });
+    return promise;
+}
+
 /**
  * { installed, groups, error } for one engine. The F5-TTS library is the plugin's own
  * input/F5-TTS folder, so it is listed even when the F5-TTS node is not installed.

web/modules/locales.js

@@ -1368,7 +1368,101 @@ export const i18n = {
         ttsCharacterError: "这个角色读取失败：{error}",
         ttsMainMissing: "没有找到合适的主声音（3~10 秒的参考音频）。点“编辑情绪”选一条。",
         ttsEmptyTitle: "还没有 GPT-SoVITS 角色",
-        ttsEmptyDesc: "把角色文件夹（里面有 .ckpt 和 .pth）放进 ComfyUI/models/gpt_sovits/，或在 extra_model_paths.yaml 里加上 gpt_sovits 目录，然后点“刷新音色”。",
+        ttsEmptyDesc: "点上面的“导入角色”，或者直接把 GPT（.ckpt）、SoVITS（.pth）和参考音频拖到这里。已经整理好的角色文件夹可以用“添加角色库”。",
+        ttsSetupTitle: "GPT-SoVITS 准备状态",
+        ttsSetupCharacters: "{count} 个角色",
+        ttsSetupPretrainedOk: "底模齐全",
+        ttsSetupPretrainedMissing: "缺 {count} 个底模",
+        ttsSetupDownloading: "底模下载中…",
+        ttsSetupPackagesMissing: "缺 {count} 个 Python 包",
+        ttsSetupLibraries: "角色库",
+        ttsSetupPretrained: "底模",
+        ttsSetupPackages: "Python 包",
+        ttsSetupRemote: "这些操作只能在运行 ComfyUI 的那台电脑上进行（当前是从别的设备打开的）。",
+        ttsSetupFailed: "操作失败：{error}",
+        ttsLibrarySource_default: "默认",
+        ttsLibrarySource_yaml: "yaml 配置",
+        ttsLibrarySource_app: "在这里添加",
+        ttsLibraryMissing: "文件夹不存在",
+        ttsLibraryAdd: "+ 添加角色库",
+        ttsLibraryAddHint: "选一个已经整理好的文件夹：里面每个子文件夹是一个角色（有 .ckpt 和 .pth）。不会复制或移动任何文件。",
+        ttsLibraryHint: "角色库里每个子文件夹是一个角色。添加后马上生效，不用重启，也不用改 extra_model_paths.yaml。",
+        ttsLibraryRemove: "移除",
+        ttsLibraryRemoveConfirm: "不再使用这个角色库？\n{path}\n\n文件不会被删除，只是不再列出里面的角色。",
+        ttsNeededFor_all: "所有模型",
+        ttsNeededFor_zh: "中文",
+        ttsNeededFor_en: "英语",
+        ttsNeededFor_ja: "日语",
+        ttsNeededFor_v2pro: "v2Pro 模型",
+        ttsPretrainedOk: "已就绪",
+        ttsPretrainedMissing: "缺少",
+        ttsPretrainedOptional: "缺少（可选）",
+        ttsPretrainedQueued: "等待下载",
+        ttsPretrainedDownloading: "下载中 {percent}%",
+        ttsPretrainedError: "下载失败",
+        ttsPretrainedDownload: "下载（约 {size}）",
+        ttsPretrainedDownloadAll: "全部下载（约 {size}）",
+        ttsPretrainedUsePackage: "使用已有的 GPT-SoVITS 整合包",
+        ttsPretrainedPackageHint: "选择整合包里的 GPT_SoVITS 文件夹（里面有 pretrained_models 和 text），就不用再下载底模了。",
+        ttsPretrainedSource: "底模来源",
+        ttsPretrainedHint: "底模也会在第一次生成时自动下载；只说日语的角色不需要中文底模。标“可选”的缺了也能用，只是效果差一点。",
+        ttsPackagesFor: "{lang}缺少 {packages}",
+        ttsPackagesHint: "在命令行运行上面的命令，然后重启 ComfyUI。",
+        ttsCopyCommand: "复制命令",
+        ttsCopied: "已复制",
+        ttsKind_gpt: "GPT",
+        ttsKind_sovits: "SoVITS",
+        ttsKind_audio: "音频",
+        ttsKind_text: "台词",
+        ttsPickFolderTitle: "选择文件夹",
+        ttsPickFilesTitle: "选择文件",
+        ttsPickPathPlaceholder: "输入或粘贴路径，按回车打开",
+        ttsPickUp: "上一级",
+        ttsPickGo: "打开",
+        ttsPickThisFolder: "选择这个文件夹",
+        ttsPickAddFiles: "添加选中的文件（{count}）",
+        ttsPickEmpty: "这里没有可用的文件或文件夹",
+        ttsPickTruncated: "文件太多，只显示了前一部分",
+        ttsImportOpen: "+ 导入角色",
+        ttsImportAddOpen: "添加文件",
+        ttsImportTitle: "导入 GPT-SoVITS 角色",
+        ttsImportSubtitle: "放入 GPT 权重（.ckpt）、SoVITS 权重（.pth）、参考音频和台词文件，自动建好角色文件夹。所有文件都是复制，原文件不动。",
+        ttsImportAddTitle: "添加文件 · {name}",
+        ttsImportAddSubtitle: "给这个角色加参考音频、情绪或新的权重。文件会复制进角色文件夹，已有的文件不会被覆盖。",
+        ttsImportName: "角色名",
+        ttsImportNamePlaceholder: "例如 阿罗娜",
+        ttsImportLibrary: "放到哪个角色库",
+        ttsImportLibraryOption: "{path}（{count} 个角色）",
+        ttsImportDropHint: "把文件拖到这里（可以一次拖多个）",
+        ttsImportDropFormats: ".ckpt · .pth · WAV / FLAC / OGG / MP3 · .txt / .list（GPT-SoVITS 标注文件，用来自动找台词）",
+        ttsImportPickFiles: "选择文件…",
+        ttsImportPickLocal: "从本机路径选择…",
+        ttsImportPickLocalHint: "直接选择电脑上的文件（比如整合包里的 GPT_weights_v2 / SoVITS_weights_v2），不用上传，更快。",
+        ttsImportSkipped: "跳过了不支持或重复的文件：{files}",
+        ttsImportRowUploading: "上传中 {percent}%",
+        ttsImportRowReady: "就绪",
+        ttsImportRowError: "失败：{error}",
+        ttsImportSeconds: "{seconds} 秒",
+        ttsImportUnsupported: "{version}（不支持）",
+        ttsImportMain: "主声音",
+        ttsImportMainTitle: "设为主声音（剧本里没标情绪的部分用它）",
+        ttsImportEmotionPlaceholder: "情绪名（可选）",
+        ttsImportTextPlaceholder: "参考台词（没找到就手动粘贴）",
+        ttsImportLanguage: "角色说的语言",
+        ttsImportLanguageAuto: "自动判断",
+        ttsImportLanguageAutoFound: "自动（识别为{lang}）",
+        ttsImportNote: "会建立：{path}",
+        ttsImportAddNote: "文件会加进「{name}」的角色文件夹。",
+        ttsImportSubmit: "导入",
+        ttsImportAddSubmit: "添加",
+        ttsImportFailed: "导入失败：{error}",
+        ttsImportNoLibrary: "没有可以写入的角色库。先在准备状态里添加一个角色库。",
+        ttsImportNoFiles: "还没有放入文件",
+        ttsImportStillUploading: "还有 {count} 个文件在上传，请稍等",
+        ttsImportNameMissing: "请填写角色名",
+        ttsImportWeightsMissing: "新角色至少要有一个 GPT 权重（.ckpt）和一个 SoVITS 权重（.pth）",
+        ttsDropNew: "松开：导入为新角色",
+        ttsDropAddTo: "松开：添加到「{name}」",
         scriptDirectorMainMissingGptSovits: "这个角色还没有主声音，请先在卡片上点“编辑情绪”选一条参考音频",
         scriptDirectorNoCharactersGptSovits: "还没有 GPT-SoVITS 角色。",
         audioDragCardHint: "拖标题栏 → F5-TTS 节点 · 拖单条 → Load Audio · 点 {标签} 复制",
@@ -2868,7 +2962,101 @@ export const i18n = {
         ttsCharacterError: "This character could not be read: {error}",
         ttsMainMissing: "No suitable main voice (a 3–10 s reference clip) was found. Use \"Edit emotions\" to pick one.",
         ttsEmptyTitle: "No GPT-SoVITS characters yet",
-        ttsEmptyDesc: "Put character folders (with .ckpt and .pth) into ComfyUI/models/gpt_sovits/, or add a gpt_sovits folder in extra_model_paths.yaml, then press Refresh Voices.",
+        ttsEmptyDesc: "Press \"Import character\" above, or drop the GPT (.ckpt), SoVITS (.pth) and reference clips here. Folders you already organised can be added with \"Add library\".",
+        ttsSetupTitle: "GPT-SoVITS setup",
+        ttsSetupCharacters: "{count} characters",
+        ttsSetupPretrainedOk: "pretrained files ready",
+        ttsSetupPretrainedMissing: "{count} pretrained files missing",
+        ttsSetupDownloading: "downloading pretrained files…",
+        ttsSetupPackagesMissing: "{count} Python packages missing",
+        ttsSetupLibraries: "Character libraries",
+        ttsSetupPretrained: "Pretrained files",
+        ttsSetupPackages: "Python packages",
+        ttsSetupRemote: "These actions only work on the computer running ComfyUI (this page was opened from another device).",
+        ttsSetupFailed: "Could not do that: {error}",
+        ttsLibrarySource_default: "default",
+        ttsLibrarySource_yaml: "yaml config",
+        ttsLibrarySource_app: "added here",
+        ttsLibraryMissing: "folder not found",
+        ttsLibraryAdd: "+ Add library",
+        ttsLibraryAddHint: "Pick a folder you already organised: each subfolder is one character (with .ckpt and .pth). Nothing is copied or moved.",
+        ttsLibraryHint: "Each subfolder of a library is one character. Added libraries work at once: no restart, no extra_model_paths.yaml.",
+        ttsLibraryRemove: "Remove",
+        ttsLibraryRemoveConfirm: "Stop using this library?\n{path}\n\nNo files are deleted; its characters are just no longer listed.",
+        ttsNeededFor_all: "all models",
+        ttsNeededFor_zh: "Chinese",
+        ttsNeededFor_en: "English",
+        ttsNeededFor_ja: "Japanese",
+        ttsNeededFor_v2pro: "v2Pro models",
+        ttsPretrainedOk: "ready",
+        ttsPretrainedMissing: "missing",
+        ttsPretrainedOptional: "missing (optional)",
+        ttsPretrainedQueued: "queued",
+        ttsPretrainedDownloading: "downloading {percent}%",
+        ttsPretrainedError: "download failed",
+        ttsPretrainedDownload: "Download (~{size})",
+        ttsPretrainedDownloadAll: "Download all (~{size})",
+        ttsPretrainedUsePackage: "Use an existing GPT-SoVITS package",
+        ttsPretrainedPackageHint: "Pick the package's GPT_SoVITS folder (it has pretrained_models and text); nothing needs downloading then.",
+        ttsPretrainedSource: "pretrained source",
+        ttsPretrainedHint: "Pretrained files are also downloaded on first use; Japanese-only characters do not need the Chinese ones. \"Optional\" ones only improve quality.",
+        ttsPackagesFor: "{lang}: missing {packages}",
+        ttsPackagesHint: "Run the command above in a terminal, then restart ComfyUI.",
+        ttsCopyCommand: "Copy command",
+        ttsCopied: "Copied",
+        ttsKind_gpt: "GPT",
+        ttsKind_sovits: "SoVITS",
+        ttsKind_audio: "audio",
+        ttsKind_text: "text",
+        ttsPickFolderTitle: "Choose a folder",
+        ttsPickFilesTitle: "Choose files",
+        ttsPickPathPlaceholder: "Type or paste a path, press Enter",
+        ttsPickUp: "Up",
+        ttsPickGo: "Open",
+        ttsPickThisFolder: "Choose this folder",
+        ttsPickAddFiles: "Add selected files ({count})",
+        ttsPickEmpty: "No usable files or folders here",
+        ttsPickTruncated: "Too many files: only the first ones are shown",
+        ttsImportOpen: "+ Import character",
+        ttsImportAddOpen: "Add files",
+        ttsImportTitle: "Import a GPT-SoVITS character",
+        ttsImportSubtitle: "Add the GPT weight (.ckpt), SoVITS weight (.pth), reference clips and line files; the character folder is built for you. Everything is copied; your files stay where they are.",
+        ttsImportAddTitle: "Add files · {name}",
+        ttsImportAddSubtitle: "Add reference clips, emotions or new weights to this character. Files are copied into its folder; existing files are never overwritten.",
+        ttsImportName: "Character name",
+        ttsImportNamePlaceholder: "e.g. Arona",
+        ttsImportLibrary: "Library",
+        ttsImportLibraryOption: "{path} ({count} characters)",
+        ttsImportDropHint: "Drop files here (several at once is fine)",
+        ttsImportDropFormats: ".ckpt · .pth · WAV / FLAC / OGG / MP3 · .txt / .list (GPT-SoVITS annotation files, used to find the lines)",
+        ttsImportPickFiles: "Choose files…",
+        ttsImportPickLocal: "Choose from this computer…",
+        ttsImportPickLocalHint: "Pick files on the computer directly (e.g. a package's GPT_weights_v2 / SoVITS_weights_v2): no upload, faster.",
+        ttsImportSkipped: "Skipped unsupported or duplicate files: {files}",
+        ttsImportRowUploading: "uploading {percent}%",
+        ttsImportRowReady: "ready",
+        ttsImportRowError: "failed: {error}",
+        ttsImportSeconds: "{seconds} s",
+        ttsImportUnsupported: "{version} (not supported)",
+        ttsImportMain: "Main voice",
+        ttsImportMainTitle: "Use as the main voice (for text without an emotion tag)",
+        ttsImportEmotionPlaceholder: "Emotion (optional)",
+        ttsImportTextPlaceholder: "Reference line (paste it if not found)",
+        ttsImportLanguage: "Character's language",
+        ttsImportLanguageAuto: "Detect",
+        ttsImportLanguageAutoFound: "Detect (found {lang})",
+        ttsImportNote: "Creates: {path}",
+        ttsImportAddNote: "Files are added to the folder of \"{name}\".",
+        ttsImportSubmit: "Import",
+        ttsImportAddSubmit: "Add",
+        ttsImportFailed: "Import failed: {error}",
+        ttsImportNoLibrary: "There is no writable character library. Add one in the setup card first.",
+        ttsImportNoFiles: "No files added yet",
+        ttsImportStillUploading: "{count} files are still uploading, please wait",
+        ttsImportNameMissing: "Enter a character name",
+        ttsImportWeightsMissing: "A new character needs at least one GPT weight (.ckpt) and one SoVITS weight (.pth)",
+        ttsDropNew: "Drop to import as a new character",
+        ttsDropAddTo: "Drop to add to \"{name}\"",
         scriptDirectorMainMissingGptSovits: "This character has no main voice yet. Pick a reference clip with \"Edit emotions\" on its card first",
         scriptDirectorNoCharactersGptSovits: "No GPT-SoVITS characters yet.",
         audioDragCardHint: "Drag the header → F5-TTS node · drag a row → Load Audio · click a {tag} to copy",

web/modules/tts_setup_api.js

@@ -0,0 +1,164 @@
+import { invalidateEngineCache } from './audio_engines.js';
+
+/**
+ * GPT-SoVITS setup and import through the Anomalous_TTS node (interface v3, §5.2–5.3).
+ * The node owns every file: character libraries, pretrained files and imported
+ * characters. This module only calls its API and holds the pure rules the import
+ * form needs; it has no DOM.
+ */
+
+export const UPLOAD_CHUNK = 8 * 1024 * 1024; // under ComfyUI's default 100 MB request limit
+
+const IMPORT_KINDS = {
+    '.ckpt': 'gpt', '.pth': 'sovits',
+    '.wav': 'audio', '.flac': 'audio', '.ogg': 'audio', '.mp3': 'audio',
+    '.txt': 'text', '.list': 'text',
+};
+const FORBIDDEN_IN_EMOTION = /[{}[\]]/;
+
+/** gpt | sovits | audio | text | null — the node checks again; this only avoids uploading unusable files. */
+export function importKind(name) {
+    const dot = String(name || '').lastIndexOf('.');
+    return dot < 0 ? null : IMPORT_KINDS[String(name).slice(dot).toLowerCase()] || null;
+}
+
+async function request(url, { method = 'POST', body, raw, signal, keepalive } = {}) {
+    const headers = raw ? { 'Content-Type': 'application/octet-stream' } : body !== undefined ? { 'Content-Type': 'application/json' } : undefined;
+    const resp = await fetch(url, {
+        method, signal, keepalive, headers,
+        body: raw ?? (body !== undefined ? JSON.stringify(body) : undefined),
+    });
+    const text = await resp.text();
+    let data = null;
+    try { data = text ? JSON.parse(text) : null; } catch (_) { /* plain-text error message */ }
+    if (!resp.ok) {
+        const error = new Error(data?.error || text || `HTTP ${resp.status}`);
+        error.status = resp.status;
+        error.data = data;
+        throw error;
+    }
+    return data;
+}
+
+// ---------- libraries, pretrained files ----------
+
+/** Add (or remove) a character library. Returns the new status. */
+export async function changeLibrary(path, remove = false) {
+    const status = await request('/anomalous_tts/libraries', { body: { path, remove } });
+    invalidateEngineCache();
+    return status;
+}
+
+/** Use (or stop using) a GPT-SoVITS package as a pretrained source. Returns the new status. */
+export async function changePretrainedSource(path, remove = false) {
+    const status = await request('/anomalous_tts/pretrained/source', { body: { path, remove } });
+    invalidateEngineCache();
+    return status;
+}
+
+/** Start background downloads; progress shows up in the status. */
+export async function startPretrainedDownload(ids) {
+    await request('/anomalous_tts/pretrained/download', { body: { ids } });
+    invalidateEngineCache();
+}
+
+export function browseFolder(path, signal) {
+    return request(`/anomalous_tts/browse${path ? `?path=${encodeURIComponent(path)}` : ''}`, { method: 'GET', signal });
+}
+
+// ---------- import ----------
+
+/**
+ * Upload one browser File in chunks. `onStart(id)` gets the upload id at once so the
+ * caller can discard it if the form is closed; a 409 resumes from what the node has.
+ */
+export async function uploadFile(file, { library, signal, onStart, onProgress } = {}) {
+    const { upload } = await request('/anomalous_tts/import/upload', { body: { name: file.name, size: file.size, library }, signal });
+    onStart?.(upload);
+    let offset = 0;
+    while (offset < file.size) {
+        try {
+            const { received } = await request(`/anomalous_tts/import/upload?upload=${upload}&offset=${offset}`, {
+                raw: file.slice(offset, offset + UPLOAD_CHUNK), signal,
+            });
+            offset = received;
+        } catch (e) {
+            if (e.status !== 409 || !Number.isInteger(e.data?.received)) throw e;
+            offset = e.data.received;
+        }
+        onProgress?.(offset / file.size);
+    }
+    return upload;
+}
+
+export function inspectImport(files, signal) {
+    return request('/anomalous_tts/import/inspect', { body: { files }, signal });
+}
+
+/** Not tied to the form's AbortSignal: closing the form must not pretend to cancel a write. */
+export async function commitImport(body) {
+    const result = await request('/anomalous_tts/import/commit', { body });
+    invalidateEngineCache();
+    return result.character;
+}
+
+/** Fire and forget; `keepalive` lets it finish while the page is closing. */
+export function discardUploads(ids) {
+    if (!ids.length) return;
+    request('/anomalous_tts/import/discard', { body: { uploads: ids }, keepalive: true }).catch(() => {});
+}
+
+/**
+ * The commit request for the import form. `rows` are in file order:
+ * { spec: {upload}|{path}, kind, emotion, text }. `language` '' = let the node decide.
+ */
+export function buildImportBody({ target = null, library = '', character = '', rows, referenceIndex = null, language = '' }) {
+    const settings = {};
+    if (language) settings.language = language;
+    const main = referenceIndex !== null ? rows[referenceIndex] : null;
+    if (main?.kind === 'audio') {
+        settings.reference = { file: referenceIndex };
+        if (main.text) settings.reference.text = main.text;
+        if (language) settings.reference.language = language;
+    }
+    const emotions = {};
+    rows.forEach((row, index) => {
+        if (row.kind !== 'audio' || index === referenceIndex || !row.emotion) return;
+        emotions[row.emotion] = { file: index };
+        if (row.text) emotions[row.emotion].text = row.text;
+    });
+    if (Object.keys(emotions).length) settings.emotions = emotions;
+    const body = { files: rows.map(row => row.spec), settings };
+    if (target) body.target = target;
+    else Object.assign(body, { library, character });
+    return body;
+}
+
+/** First reason the form cannot be sent, as [locale key, params], or null. */
+export function importProblem({ target = null, character = '', rows, uploading = 0 }) {
+    if (!rows.length) return ['ttsImportNoFiles', {}];
+    if (uploading) return ['ttsImportStillUploading', { count: uploading }];
+    if (!target) {
+        if (!character.trim()) return ['ttsImportNameMissing', {}];
+        if (!rows.some(row => row.kind === 'gpt') || !rows.some(row => row.kind === 'sovits')) return ['ttsImportWeightsMissing', {}];
+    }
+    const seen = new Set();
+    for (const row of rows) {
+        if (row.kind !== 'audio' || !row.emotion) continue;
+        if (row.emotion === 'main' || FORBIDDEN_IN_EMOTION.test(row.emotion)) return ['ttsEditorNameInvalid', { name: row.emotion }];
+        if (seen.has(row.emotion)) return ['ttsEditorNameDuplicate', { name: row.emotion }];
+        seen.add(row.emotion);
+    }
+    return null;
+}
+
+/** Summary for the setup card: what still needs doing. */
+export function setupSummary(status) {
+    const characters = (status?.libraries || []).reduce((sum, lib) => sum + (lib.characters || 0), 0);
+    const missing = (status?.pretrained || []).filter(item => item.state !== 'ok');
+    const packages = Object.values(status?.dependencies || {}).reduce((sum, dep) => sum + (dep.missing?.length || 0), 0);
+    const downloading = missing.some(item => item.state === 'queued' || item.state === 'downloading');
+    const requiredMissing = missing.filter(item => item.required).length;
+    return { characters, missing: missing.length, requiredMissing, packages, downloading,
+        ready: characters > 0 && requiredMissing === 0 && packages === 0 };
+}

web/modules/ui_audio_studio.js

@@ -5,8 +5,10 @@ import { stopGalleryAudio } from './ui_audio_gallery.js';
 import { openAudioUploaderModal } from './ui_audio_uploader.js';
 import { getActiveAudioFilter, renderAudioSidebar, setActiveAudioFilter } from './ui_audio_sidebar.js';
 import { AUDIO_NODE_TARGETS, alignedVoiceValue, planVoiceDrop } from './audio_node_targets.js';
-import { AUDIO_ENGINES, detectEngines, engineById, engineTargetLabels, getStoredEngine, invalidateEngineCache, loadEngine, pickEngine, setStoredEngine } from './audio_engines.js';
+import { AUDIO_ENGINES, detectEngines, engineById, engineTargetLabels, getStoredEngine, invalidateEngineCache, loadEngine, loadGptSovitsStatus, pickEngine, setStoredEngine } from './audio_engines.js';
 import { openGptSovitsEditor } from './ui_audio_tts_editor.js';
+import { renderTtsSetup } from './ui_tts_setup.js';
+import { bindTtsFileDrop, openTtsImport } from './ui_tts_import.js';
 import { bindMaterialDrag } from './material_drag.js';
 import {
     openScriptDirector,
@@ -263,10 +265,11 @@ function renderSliceRow(slice, owner) {
     return row;
 }
 
-function renderCharacterCard(group, owner, { onChanged } = {}) {
+function renderCharacterCard(group, owner, { onChanged, canImport = false } = {}) {
     const isTts = group.engine === 'gpt_sovits';
     const card = document.createElement('div');
     card.className = 'anomalous-character-voice-card';
+    if (isTts && canImport && !group.error) card.dataset.ttsCharacter = group.character; // file drop target
 
     const header = document.createElement('div');
     header.className = 'anomalous-character-voice-header';
@@ -314,6 +317,11 @@ function renderCharacterCard(group, owner, { onChanged } = {}) {
         const edit = createToolButton(SVG.EDIT, t('ttsEditorOpen'), 'is-compact');
         edit.onclick = () => openGptSovitsEditor(group, { onSaved: () => onChanged?.() });
         headerRight.append(edit);
+        if (canImport) {
+            const add = createToolButton(SVG.PLUS, t('ttsImportAddOpen'), 'is-compact');
+            add.onclick = () => openTtsImport({ target: group.character, onDone: () => onChanged?.() });
+            headerRight.append(add);
+        }
     }
     headerRight.append(countBadge);
     header.append(titleGroup, headerRight);
@@ -614,7 +622,9 @@ export async function renderAudioStudio(container, { filter = null, owner = null
         studioWrapper.lastChild.replaceWith(renderInstallCard(engine));
         return;
     }
-    const result = await loadEngine(engineId);
+    // The setup status only exists for GPT-SoVITS nodes with interface v3 (null otherwise).
+    const statusRequest = engineId === 'gpt_sovits' ? loadGptSovitsStatus().catch(() => null) : null;
+    const [result, ttsStatus] = await Promise.all([loadEngine(engineId), statusRequest]);
     if (renderTokens.get(container) !== token) return;
     if (result.error) {
         studioWrapper.lastChild.replaceWith(renderStatus('anomalous-audio-status is-error', t('audioLoadFailed', { error: result.error })));
@@ -624,6 +634,14 @@ export async function renderAudioStudio(container, { filter = null, owner = null
     updateScriptDirectorVoices(characters, characterFilter?.value || null, engineId);
     if (!installed) studioWrapper.insertBefore(renderInstallCard(engine, { compact: true }), studioWrapper.lastChild);
 
+    const onChanged = () => { rerender(); if (owner) renderAudioSidebar(owner); };
+    const canImport = Boolean(ttsStatus) && ttsStatus.local !== false;
+    if (ttsStatus) {
+        const onImport = (options = {}) => openTtsImport({ ...options, onDone: onChanged });
+        studioWrapper.insertBefore(renderTtsSetup(ttsStatus, { onChanged, onImport: () => onImport() }), studioWrapper.lastChild);
+        if (canImport) bindTtsFileDrop(studioWrapper, (files, target) => onImport({ files, target }));
+    }
+
     if (characterFilter) {
         characters = characters.filter(group => group.group === characterFilter.value);
     }
@@ -632,7 +650,6 @@ export async function renderAudioStudio(container, { filter = null, owner = null
         studioWrapper.lastChild.replaceWith(renderEmptyGuide(engineId));
         return;
     }
-    const onChanged = () => { rerender(); if (owner) renderAudioSidebar(owner); };
-    characters.forEach(group => grid.appendChild(renderCharacterCard(group, owner, { onChanged })));
+    characters.forEach(group => grid.appendChild(renderCharacterCard(group, owner, { onChanged, canImport })));
     studioWrapper.lastChild.replaceWith(grid);
 }

web/modules/ui_tts_import.js

@@ -0,0 +1,422 @@
+import { t } from './interface_settings.js';
+import { createViewScope } from './ui_lifecycle.js';
+import { anomalousAlert } from './ui_dialog.js';
+import { loadGptSovitsStatus } from './audio_engines.js';
+import { buildImportBody, commitImport, discardUploads, importKind, importProblem, inspectImport, uploadFile } from './tts_setup_api.js';
+import { PICKER_OVERLAY_CLASS, formatSize, pickServerPath } from './ui_tts_path_picker.js';
+
+/**
+ * Import GPT-SoVITS files as a new character, or add them to one (`target`), through
+ * the Anomalous_TTS node. Files come from a drop, the browser's file dialog (both are
+ * uploaded in chunks) or the node's folder picker (sent as paths). The node copies
+ * everything; the originals are never touched.
+ *
+ * Each file is one row object; its DOM is built once and updated in place, so
+ * upload progress and inspect results never steal focus from a field being typed in.
+ */
+
+let activeScope = null;
+const LANGUAGES = ['', 'ja', 'zh', 'en'];
+const ACCEPT = '.ckpt,.pth,.wav,.flac,.ogg,.mp3,.txt,.list';
+
+function el(tag, className, text) {
+    const node = document.createElement(tag);
+    if (className) node.className = className;
+    if (text !== undefined) node.textContent = text;
+    return node;
+}
+
+function button(className, label, onClick, title) {
+    const btn = el('button', className, label);
+    btn.type = 'button';
+    if (title) { btn.title = title; btn.setAttribute('aria-label', title); }
+    btn.onclick = onClick;
+    return btn;
+}
+
+function input(className, placeholder, value = '') {
+    const node = el('input', className);
+    node.type = 'text';
+    node.placeholder = placeholder;
+    node.value = value;
+    return node;
+}
+
+/**
+ * Let OS files be dropped on `root` (the studio). `onDrop(files, character)` gets the
+ * character of the card they were dropped on, or null. Internal drags are ignored.
+ */
+export function bindTtsFileDrop(root, onDrop) {
+    const hasFiles = e => [...(e.dataTransfer?.types || [])].includes('Files');
+    let marked = null;
+    root.dataset.dropHint = t('ttsDropNew');
+    const mark = (card) => {
+        if (marked === card) return;
+        marked?.classList.remove('is-file-target');
+        card?.classList.add('is-file-target');
+        marked = card;
+        root.dataset.dropHint = card ? t('ttsDropAddTo', { name: card.dataset.ttsCharacter }) : t('ttsDropNew');
+    };
+    const cardAt = e => e.target?.closest?.('[data-tts-character]') || null;
+    root.addEventListener('dragover', (e) => {
+        if (!hasFiles(e)) return;
+        e.preventDefault();
+        e.dataTransfer.dropEffect = 'copy';
+        root.classList.add('is-file-dragover');
+        mark(cardAt(e));
+    });
+    root.addEventListener('dragleave', (e) => {
+        if (root.contains(e.relatedTarget)) return;
+        root.classList.remove('is-file-dragover');
+        mark(null);
+    });
+    root.addEventListener('drop', (e) => {
+        if (!hasFiles(e)) return;
+        e.preventDefault();
+        e.stopPropagation(); // ComfyUI would try to open the files as a workflow
+        const card = cardAt(e);
+        root.classList.remove('is-file-dragover');
+        mark(null);
+        onDrop([...e.dataTransfer.files], card?.dataset.ttsCharacter || null);
+    });
+}
+
+/** Open the import form. `onDone(character)` runs after the node created or updated it. */
+export async function openTtsImport({ files = [], target = null, onDone } = {}) {
+    let status;
+    try {
+        status = await loadGptSovitsStatus();
+    } catch (e) {
+        await anomalousAlert(t('ttsSetupFailed', { error: e.message || String(e) }));
+        return;
+    }
+    if (!status) return;
+    if (status.local === false) {
+        await anomalousAlert(t('ttsSetupRemote'));
+        return;
+    }
+    const libraries = status.libraries.filter(lib => lib.writable);
+    if (!target && !libraries.length) {
+        await anomalousAlert(t('ttsImportNoLibrary'));
+        return;
+    }
+
+    activeScope?.dispose();
+    const scope = createViewScope();
+    activeScope = scope;
+    const rows = [];
+    let committed = false;
+    let referenceKey = null;
+    let referenceChosen = Boolean(target); // adding files never replaces the main voice unless asked
+    let nameEdited = false;
+    let inspectTimer = null;
+    let inspectToken = 0;
+    let player = null;
+    let nextKey = 0;
+    let uploads = Promise.resolve();
+
+    const overlay = el('div', 'anomalous-voice-modal-overlay');
+    const modal = el('div', 'anomalous-voice-modal anomalous-tts-import');
+    modal.setAttribute('role', 'dialog');
+    modal.setAttribute('aria-modal', 'true');
+    const close = () => scope.dispose();
+    scope.onDispose(() => {
+        clearTimeout(inspectTimer);
+        player?.pause();
+        for (const row of rows) {
+            row.controller.abort();
+            if (row.url) URL.revokeObjectURL(row.url);
+        }
+        if (!committed) discardUploads(rows.map(row => row.uploadId).filter(Boolean));
+        overlay.remove();
+        if (activeScope === scope) activeScope = null;
+    });
+    // No close on a backdrop click: a stray click would throw away the files and uploads.
+    scope.listen(window, 'keydown', (e) => {
+        if (e.key !== 'Escape' || document.querySelector(`.anomalous-dialog-overlay, .${PICKER_OVERLAY_CLASS}`)) return;
+        e.stopPropagation();
+        close();
+    }, true);
+
+    // ---- header, target ----
+    const header = el('div', 'anomalous-voice-modal-header');
+    const heading = el('div', 'anomalous-voice-modal-heading');
+    heading.append(
+        el('div', 'anomalous-voice-modal-title', target ? t('ttsImportAddTitle', { name: target }) : t('ttsImportTitle')),
+        el('div', 'anomalous-voice-modal-subtitle', t(target ? 'ttsImportAddSubtitle' : 'ttsImportSubtitle')),
+    );
+    header.append(heading, button('anomalous-voice-modal-close', '×', close, t('close')));
+
+    const nameInput = input('anomalous-uploader-input', t('ttsImportNamePlaceholder'));
+    nameInput.oninput = () => { nameEdited = true; syncNote(); };
+    const librarySelect = el('select', 'anomalous-uploader-input');
+    for (const lib of libraries) {
+        const option = el('option', '', t('ttsImportLibraryOption', { path: lib.path, count: lib.characters }));
+        option.value = lib.path;
+        librarySelect.append(option);
+    }
+    librarySelect.onchange = () => syncNote();
+    const targetBox = el('div', 'anomalous-tts-import-target');
+    if (!target) {
+        const nameField = el('label', 'anomalous-voice-field');
+        nameField.append(el('span', 'anomalous-voice-field-label', t('ttsImportName')), nameInput);
+        const libField = el('label', 'anomalous-voice-field');
+        libField.append(el('span', 'anomalous-voice-field-label', t('ttsImportLibrary')), librarySelect);
+        targetBox.append(nameField, libField);
+    }
+
+    // ---- sources ----
+    const fileInput = el('input');
+    fileInput.type = 'file';
+    fileInput.multiple = true;
+    fileInput.accept = ACCEPT;
+    fileInput.hidden = true;
+    fileInput.onchange = () => { addBrowserFiles([...fileInput.files]); fileInput.value = ''; };
+    const dropzone = el('div', 'anomalous-audio-dropzone anomalous-tts-import-drop');
+    const pickLocal = button('anomalous-tts-add', t('ttsImportPickLocal'), async (e) => {
+        e.stopPropagation();
+        const paths = await pickServerPath({ mode: 'files', title: t('ttsImportPickLocal'), hint: t('ttsImportPickLocalHint') });
+        if (paths && !scope.signal.aborted) addPaths(paths);
+    });
+    const pickBrowser = button('anomalous-tts-add', t('ttsImportPickFiles'), (e) => { e.stopPropagation(); fileInput.click(); });
+    const dropButtons = el('div', 'anomalous-tts-setup-actions');
+    dropButtons.append(pickBrowser, pickLocal);
+    dropzone.append(el('div', 'anomalous-audio-dropzone-hint', t('ttsImportDropHint')),
+        el('div', 'anomalous-audio-dropzone-formats', t('ttsImportDropFormats')), dropButtons, fileInput);
+    dropzone.ondragover = (e) => { e.preventDefault(); dropzone.classList.add('dragover'); };
+    dropzone.ondragleave = () => dropzone.classList.remove('dragover');
+    dropzone.ondrop = (e) => {
+        e.preventDefault();
+        e.stopPropagation();
+        dropzone.classList.remove('dragover');
+        addBrowserFiles([...(e.dataTransfer?.files || [])]);
+    };
+
+    const list = el('div', 'anomalous-tts-import-list');
+    const skipped = el('div', 'anomalous-tts-hint');
+    const problems = el('div', 'anomalous-tts-import-problems');
+
+    const languageSelect = el('select', 'anomalous-uploader-input');
+    const languageField = el('label', 'anomalous-voice-field anomalous-tts-import-language');
+    languageField.append(el('span', 'anomalous-voice-field-label', t('ttsImportLanguage')), languageSelect);
+    const fillLanguages = (suggested) => {
+        const keep = languageSelect.value;
+        languageSelect.replaceChildren(...LANGUAGES.map(code => {
+            const option = el('option', '', code ? t(`ttsNeededFor_${code}`)
+                : suggested ? t('ttsImportLanguageAutoFound', { lang: t(`ttsNeededFor_${suggested}`) }) : t('ttsImportLanguageAuto'));
+            option.value = code;
+            return option;
+        }));
+        languageSelect.value = keep;
+    };
+    fillLanguages(null);
+    const note = el('div', 'anomalous-tts-hint');
+    const syncNote = () => {
+        note.textContent = target ? t('ttsImportAddNote', { name: target })
+            : t('ttsImportNote', { path: `${librarySelect.value}/${nameInput.value.trim() || '…'}` });
+    };
+    syncNote();
+
+    // ---- rows ----
+    const readyRows = () => rows.filter(row => row.spec && !row.error);
+
+    function renderState(row) {
+        const { state } = row.els;
+        state.classList.toggle('is-error', Boolean(row.error));
+        state.textContent = row.error ? t('ttsImportRowError', { error: row.error })
+            : row.spec ? t('ttsImportRowReady') : t('ttsImportRowUploading', { percent: Math.floor(100 * row.progress) });
+    }
+
+    function renderMeta(row) {
+        const info = row.info || {};
+        const parts = row.size ? [formatSize(row.size)] : [];
+        if (info.version) parts.push(info.supported ? info.version : t('ttsImportUnsupported', { version: info.version }));
+        if (info.seconds !== undefined) parts.push(t('ttsImportSeconds', { seconds: info.seconds }));
+        row.els.meta.textContent = parts.join(' · ');
+    }
+
+    function syncReference() {
+        for (const row of rows) {
+            if (!row.els.main) continue;
+            const on = row.key === referenceKey;
+            row.els.main.setAttribute('aria-pressed', String(on));
+            row.els.main.classList.toggle('active', on);
+            row.els.emotion.disabled = on;
+        }
+    }
+
+    function removeRow(row) {
+        row.controller.abort();
+        if (row.uploadId) discardUploads([row.uploadId]);
+        if (row.url) URL.revokeObjectURL(row.url);
+        rows.splice(rows.indexOf(row), 1);
+        row.els.root.remove();
+        if (referenceKey === row.key) referenceKey = null;
+        scheduleInspect();
+    }
+
+    function buildRow(row) {
+        const root = el('div', 'anomalous-tts-import-row');
+        root.dataset.kind = row.kind;
+        const line = el('div', 'anomalous-tts-import-line');
+        const name = el('span', 'anomalous-tts-picker-name', row.name);
+        name.title = row.path || row.name;
+        row.els = { root, meta: el('span', 'anomalous-tts-picker-meta'), state: el('span', 'anomalous-tts-import-state') };
+        line.append(el('span', 'anomalous-tts-kind', t(`ttsKind_${row.kind}`)), name, row.els.meta, row.els.state,
+            button('anomalous-tts-remove', '×', () => removeRow(row), t('ttsEditorRemove')));
+        root.append(line);
+        if (row.kind === 'audio') {
+            const audio = el('div', 'anomalous-tts-import-audio');
+            row.els.main = button('anomalous-tag-chip', t('ttsImportMain'), () => {
+                referenceKey = referenceKey === row.key ? null : row.key;
+                referenceChosen = true;
+                syncReference();
+            }, t('ttsImportMainTitle'));
+            const play = button('anomalous-audio-play-btn', '▶', () => {
+                player?.pause();
+                row.url = row.url || URL.createObjectURL(row.file);
+                player = new Audio(row.url);
+                player.play().catch(() => {});
+            }, t('audioPlay'));
+            play.hidden = !row.file; // local paths cannot be played from the browser
+            row.els.emotion = input('anomalous-tts-name-input', t('ttsImportEmotionPlaceholder'));
+            row.els.emotion.oninput = () => { row.emotion = row.els.emotion.value.trim(); };
+            row.els.text = input('anomalous-tts-text-input', t('ttsImportTextPlaceholder'));
+            row.els.text.oninput = () => { row.text = row.els.text.value.trim(); row.textEdited = true; };
+            audio.append(row.els.main, play, row.els.emotion, row.els.text);
+            root.append(audio);
+        }
+        renderMeta(row);
+        renderState(row);
+        return root;
+    }
+
+    function addRow(fields) {
+        const duplicate = rows.some(row => row.kind === fields.kind && row.name.toLowerCase() === fields.name.toLowerCase());
+        if (duplicate) return null;
+        const row = { key: nextKey++, progress: 0, emotion: '', text: '', textEdited: false, uploadId: null,
+            spec: null, error: '', info: null, controller: new AbortController(), ...fields };
+        rows.push(row);
+        list.append(buildRow(row));
+        return row;
+    }
+
+    function reportSkipped(names) {
+        skipped.textContent = names.length ? t('ttsImportSkipped', { files: names.join(', ') }) : '';
+    }
+
+    function addBrowserFiles(fileList) {
+        const bad = [];
+        for (const file of fileList) {
+            const kind = importKind(file.name);
+            const row = kind ? addRow({ name: file.name, kind, size: file.size, file }) : null;
+            if (!row) { bad.push(file.name); continue; }
+            uploads = uploads.then(() => upload(row));
+        }
+        reportSkipped(bad);
+    }
+
+    function addPaths(paths) {
+        const bad = [];
+        for (const path of paths) {
+            const name = path.split('/').pop();
+            const kind = importKind(name);
+            const row = kind ? addRow({ name, kind, size: 0, path, spec: { path } }) : null;
+            if (!row) bad.push(name);
+        }
+        reportSkipped(bad);
+        scheduleInspect();
+    }
+
+    async function upload(row) {
+        if (row.controller.signal.aborted) return;
+        try {
+            await uploadFile(row.file, {
+                library: target ? undefined : librarySelect.value,
+                signal: row.controller.signal,
+                onStart: (id) => { row.uploadId = id; },
+                onProgress: (fraction) => { row.progress = fraction; renderState(row); },
+            });
+            row.spec = { upload: row.uploadId };
+        } catch (e) {
+            if (row.controller.signal.aborted) return;
+            row.error = e.message || String(e);
+        }
+        renderState(row);
+        scheduleInspect();
+    }
+
+    function scheduleInspect() {
+        clearTimeout(inspectTimer);
+        inspectTimer = setTimeout(runInspect, 250);
+    }
+
+    async function runInspect() {
+        const ready = readyRows();
+        if (!ready.length || rows.some(row => !row.spec && !row.error)) {
+            if (!ready.length) problems.replaceChildren();
+            return;
+        }
+        const mine = ++inspectToken;
+        try {
+            const result = await inspectImport(ready.map(row => row.spec), scope.signal);
+            if (mine !== inspectToken || scope.signal.aborted) return;
+            ready.forEach((row, i) => {
+                row.info = result.files[i];
+                if (row.info.size) row.size = row.info.size;
+                renderMeta(row);
+                if (row.els.text && !row.textEdited && row.info.text) {
+                    row.text = row.info.text;
+                    row.els.text.value = row.info.text;
+                }
+            });
+            const { suggested } = result;
+            if (!target && !nameEdited && suggested.name) { nameInput.value = suggested.name; syncNote(); }
+            if (!referenceChosen) referenceKey = suggested.reference !== null ? ready[suggested.reference]?.key ?? null : null;
+            syncReference();
+            fillLanguages(suggested.language);
+            problems.replaceChildren(...result.problems.map(text => el('div', 'anomalous-tts-import-problem', text)));
+        } catch (e) {
+            if (mine !== inspectToken || scope.signal.aborted) return;
+            problems.replaceChildren(el('div', 'anomalous-tts-import-problem is-error', e.message || String(e)));
+        }
+    }
+
+    // ---- footer ----
+    const footer = el('div', 'anomalous-voice-modal-footer');
+    const submit = button('anomalous-voice-modal-submit', t(target ? 'ttsImportAddSubmit' : 'ttsImportSubmit'), async () => {
+        const ready = readyRows();
+        const pending = rows.filter(row => !row.spec && !row.error).length;
+        const problem = importProblem({ target, character: nameInput.value, rows: ready, uploading: pending });
+        if (problem) {
+            await anomalousAlert(t(problem[0], problem[1]));
+            return;
+        }
+        const index = ready.findIndex(row => row.key === referenceKey);
+        const body = buildImportBody({
+            target, library: librarySelect.value, character: nameInput.value.trim(), language: languageSelect.value,
+            referenceIndex: index >= 0 ? index : null,
+            rows: ready.map(row => ({ spec: row.spec, kind: row.kind, emotion: row.emotion, text: row.text })),
+        });
+        submit.disabled = true;
+        try {
+            const character = await commitImport(body);
+            committed = true;
+            onDone?.(character);
+            close();
+        } catch (e) {
+            if (!scope.signal.aborted) await anomalousAlert(t('ttsImportFailed', { error: e.message || String(e) }));
+        } finally {
+            if (!scope.signal.aborted) submit.disabled = false;
+        }
+    });
+    footer.append(button('anomalous-voice-modal-cancel', t('dialogCancel'), close), submit);
+
+    const body = el('div', 'anomalous-tts-editor-body');
+    body.append(targetBox, dropzone, skipped, list, languageField, problems, note);
+    modal.append(header, body, footer);
+    overlay.appendChild(modal);
+    document.body.appendChild(overlay);
+    addBrowserFiles(files);
+    (target ? pickBrowser : nameInput).focus();
+}

web/modules/ui_tts_path_picker.js

@@ -0,0 +1,159 @@
+import { t } from './interface_settings.js';
+import { createViewScope } from './ui_lifecycle.js';
+import { browseFolder } from './tts_setup_api.js';
+
+/**
+ * Choose a folder, or files, on the computer running ComfyUI. The browser cannot
+ * see local paths, so the Anomalous_TTS node lists folders (`/anomalous_tts/browse`).
+ * Resolves with the folder path, an array of file paths, or null when cancelled.
+ */
+
+const LAST_PATH_KEY = 'anomalous_tts_last_path';
+export const PICKER_OVERLAY_CLASS = 'anomalous-tts-picker-overlay';
+const FOLDER_SVG = '<svg width="13" height="13" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true"><path d="M3 7a2 2 0 0 1 2-2h4l2 2h8a2 2 0 0 1 2 2v8a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2Z"/></svg>';
+
+function el(tag, className, text) {
+    const node = document.createElement(tag);
+    if (className) node.className = className;
+    if (text !== undefined) node.textContent = text;
+    return node;
+}
+
+function button(className, label, onClick, title) {
+    const btn = el('button', className, label);
+    btn.type = 'button';
+    if (title) { btn.title = title; btn.setAttribute('aria-label', title); }
+    btn.onclick = onClick;
+    return btn;
+}
+
+function lastPath() {
+    try { return localStorage.getItem(LAST_PATH_KEY) || ''; } catch (_) { return ''; }
+}
+
+function rememberPath(path) {
+    try { localStorage.setItem(LAST_PATH_KEY, path); } catch (_) { /* convenience only */ }
+}
+
+export function formatSize(bytes) {
+    if (bytes >= 1024 ** 3) return `${(bytes / 1024 ** 3).toFixed(1)} GB`;
+    if (bytes >= 1024 ** 2) return `${(bytes / 1024 ** 2).toFixed(1)} MB`;
+    return `${Math.max(1, Math.round(bytes / 1024))} KB`;
+}
+
+/**
+ * `mode`: 'folder' (resolve with the open folder) or 'files' (tick files, resolve with their paths).
+ * `kinds` limits the files shown in 'files' mode (gpt, sovits, audio, text).
+ */
+export function pickServerPath({ mode = 'folder', title, hint = '', kinds = null } = {}) {
+    return new Promise(resolve => {
+        const scope = createViewScope();
+        let result = null;
+        let current = '';
+        let token = 0;
+        const picked = new Set();
+
+        const overlay = el('div', `anomalous-voice-modal-overlay ${PICKER_OVERLAY_CLASS}`);
+        const modal = el('div', 'anomalous-voice-modal anomalous-tts-picker');
+        modal.setAttribute('role', 'dialog');
+        modal.setAttribute('aria-modal', 'true');
+        const close = () => scope.dispose();
+        scope.onDispose(() => { overlay.remove(); resolve(result); });
+        overlay.onclick = (e) => { if (e.target === overlay) close(); };
+        scope.listen(window, 'keydown', (e) => {
+            if (e.key !== 'Escape' || document.querySelector('.anomalous-dialog-overlay')) return;
+            e.stopPropagation();
+            close();
+        }, true);
+
+        const header = el('div', 'anomalous-voice-modal-header');
+        const heading = el('div', 'anomalous-voice-modal-heading');
+        heading.append(el('div', 'anomalous-voice-modal-title', title || t(mode === 'folder' ? 'ttsPickFolderTitle' : 'ttsPickFilesTitle')));
+        if (hint) heading.append(el('div', 'anomalous-voice-modal-subtitle', hint));
+        header.append(heading, button('anomalous-voice-modal-close', '×', close, t('close')));
+
+        const pathInput = el('input', 'anomalous-tts-picker-path');
+        pathInput.type = 'text';
+        pathInput.placeholder = t('ttsPickPathPlaceholder');
+        pathInput.onkeydown = (e) => { if (e.key === 'Enter') { e.preventDefault(); open(pathInput.value.trim()); } };
+        const upBtn = button('anomalous-tts-add', t('ttsPickUp'), () => open(upBtn.dataset.parent || ''));
+        const goBtn = button('anomalous-tts-add', t('ttsPickGo'), () => open(pathInput.value.trim()));
+        const bar = el('div', 'anomalous-tts-picker-bar');
+        bar.append(upBtn, pathInput, goBtn);
+
+        const list = el('div', 'anomalous-tts-picker-list');
+        list.setAttribute('role', 'listbox');
+        const status = el('div', 'anomalous-tts-hint');
+
+        const footer = el('div', 'anomalous-voice-modal-footer');
+        const confirm = button('anomalous-voice-modal-submit', '', () => {
+            if (mode === 'folder') {
+                if (!current) return;
+                result = current;
+            } else {
+                if (!picked.size) return;
+                result = [...picked];
+            }
+            rememberPath(current);
+            close();
+        });
+        const syncConfirm = () => {
+            confirm.textContent = mode === 'folder' ? t('ttsPickThisFolder') : t('ttsPickAddFiles', { count: picked.size });
+            confirm.disabled = mode === 'folder' ? !current : !picked.size;
+        };
+        footer.append(button('anomalous-voice-modal-cancel', t('dialogCancel'), close), confirm);
+
+        const fileRow = (folder, file) => {
+            const path = `${folder.replace(/\/$/, '')}/${file.name}`;
+            const row = el('label', 'anomalous-tts-picker-item is-file');
+            const box = el('input');
+            box.type = 'checkbox';
+            box.checked = picked.has(path);
+            box.onchange = () => { if (box.checked) picked.add(path); else picked.delete(path); syncConfirm(); };
+            row.append(box, el('span', 'anomalous-tts-picker-name', file.name),
+                el('span', 'anomalous-tts-kind', t(`ttsKind_${file.kind}`)), el('span', 'anomalous-tts-picker-meta', formatSize(file.size)));
+            return row;
+        };
+
+        async function open(path) {
+            const mine = ++token;
+            status.textContent = t('audioLoading');
+            status.classList.remove('is-error');
+            try {
+                const data = await browseFolder(path, scope.signal);
+                if (mine !== token || scope.signal.aborted) return;
+                current = data.path;
+                pathInput.value = data.path;
+                upBtn.disabled = data.parent === null;
+                upBtn.dataset.parent = data.parent || '';
+                const items = data.dirs.map(name => {
+                    const target = data.path ? `${data.path.replace(/\/$/, '')}/${name}` : name;
+                    const row = button('anomalous-tts-picker-item is-dir', '', () => open(target));
+                    const icon = el('span', 'anomalous-tts-picker-icon');
+                    icon.innerHTML = FOLDER_SVG;
+                    row.append(icon, el('span', 'anomalous-tts-picker-name', name));
+                    return row;
+                });
+                if (mode === 'files') {
+                    items.push(...data.files.filter(f => !kinds || kinds.includes(f.kind)).map(f => fileRow(data.path, f)));
+                }
+                list.replaceChildren(...items);
+                status.textContent = !items.length ? t('ttsPickEmpty') : data.truncated ? t('ttsPickTruncated') : '';
+                syncConfirm();
+            } catch (e) {
+                if (mine !== token || scope.signal.aborted) return;
+                if (path && path === initial && !current) { open(''); return; } // remembered folder is gone
+                status.textContent = e.message || String(e);
+                status.classList.add('is-error');
+            }
+        }
+
+        modal.append(header, bar, list, status, footer);
+        overlay.appendChild(modal);
+        document.body.appendChild(overlay);
+        syncConfirm();
+        const initial = lastPath();
+        open(initial);
+        pathInput.focus();
+    });
+}

web/modules/ui_tts_setup.js

@@ -0,0 +1,218 @@
+import { t } from './interface_settings.js';
+import { anomalousAlert, anomalousConfirm } from './ui_dialog.js';
+import { loadGptSovitsStatus } from './audio_engines.js';
+import { changeLibrary, changePretrainedSource, setupSummary, startPretrainedDownload } from './tts_setup_api.js';
+import { formatSize, pickServerPath } from './ui_tts_path_picker.js';
+
+/**
+ * GPT-SoVITS setup card at the top of the audio studio: character libraries,
+ * pretrained files (download or use an existing GPT-SoVITS package) and missing
+ * Python packages, from the node's `/anomalous_tts/status`. Collapsed to one line
+ * once everything is ready. While a download runs the card polls the status and
+ * redraws itself; it stops as soon as it is no longer in the page.
+ */
+
+const OPEN_KEY = 'anomalous_tts_setup_open';
+const POLL_MS = 1000;
+
+function el(tag, className, text) {
+    const node = document.createElement(tag);
+    if (className) node.className = className;
+    if (text !== undefined) node.textContent = text;
+    return node;
+}
+
+function button(className, label, onClick, title) {
+    const btn = el('button', className, label);
+    btn.type = 'button';
+    if (title) btn.title = title;
+    btn.onclick = onClick;
+    return btn;
+}
+
+function storedOpen() {
+    try { return localStorage.getItem(OPEN_KEY); } catch (_) { return null; }
+}
+
+function storeOpen(open) {
+    try { localStorage.setItem(OPEN_KEY, open ? '1' : '0'); } catch (_) { /* convenience only */ }
+}
+
+/**
+ * Run a change, report failures, then let the studio redraw. The button stays disabled
+ * while it runs; `work` returns false when the user cancelled (nothing to redraw).
+ */
+async function act(btn, work, onChanged) {
+    btn.disabled = true;
+    try {
+        if (await work() === false) btn.disabled = false;
+        else onChanged();
+    } catch (e) {
+        await anomalousAlert(t('ttsSetupFailed', { error: e.message || String(e) }));
+        btn.disabled = false;
+    }
+}
+
+function summaryText(status, summary) {
+    const parts = [t('ttsSetupCharacters', { count: summary.characters })];
+    parts.push(summary.downloading ? t('ttsSetupDownloading')
+        : summary.missing ? t('ttsSetupPretrainedMissing', { count: summary.missing }) : t('ttsSetupPretrainedOk'));
+    if (summary.packages) parts.push(t('ttsSetupPackagesMissing', { count: summary.packages }));
+    return parts.join(' · ');
+}
+
+function renderLibraries(status, local, onChanged) {
+    const group = el('div', 'anomalous-tts-setup-group');
+    group.append(el('div', 'anomalous-tts-section', t('ttsSetupLibraries')));
+    for (const lib of status.libraries) {
+        const row = el('div', 'anomalous-tts-setup-row');
+        const path = el('span', 'anomalous-tts-setup-path', lib.path);
+        path.title = lib.path;
+        row.append(path, el('span', 'anomalous-tts-kind', t(`ttsLibrarySource_${lib.source}`)),
+            el('span', 'anomalous-tts-setup-meta', lib.exists ? t('ttsSetupCharacters', { count: lib.characters }) : t('ttsLibraryMissing')));
+        if (lib.source === 'app') {
+            const remove = button('anomalous-tts-remove', '×', () => act(remove, async () => {
+                if (!await anomalousConfirm(t('ttsLibraryRemoveConfirm', { path: lib.path }))) return false;
+                await changeLibrary(lib.path, true);
+            }, onChanged), t('ttsLibraryRemove'));
+            remove.disabled = !local;
+            row.append(remove);
+        }
+        group.append(row);
+    }
+    const add = button('anomalous-tts-add', t('ttsLibraryAdd'), () => act(add, async () => {
+        const path = await pickServerPath({ mode: 'folder', title: t('ttsLibraryAdd'), hint: t('ttsLibraryAddHint') });
+        if (!path) return false;
+        await changeLibrary(path);
+    }, onChanged));
+    add.disabled = !local;
+    group.append(add, el('div', 'anomalous-tts-hint', t('ttsLibraryHint')));
+    return group;
+}
+
+function pretrainedState(item) {
+    if (item.state === 'ok') return el('span', 'anomalous-tts-state is-ok', t('ttsPretrainedOk'));
+    if (item.state === 'downloading') {
+        const pct = item.size ? Math.min(99, Math.floor(100 * (item.done || 0) / item.size)) : 0;
+        return el('span', 'anomalous-tts-state is-busy', t('ttsPretrainedDownloading', { percent: pct }));
+    }
+    if (item.state === 'queued') return el('span', 'anomalous-tts-state is-busy', t('ttsPretrainedQueued'));
+    const state = el('span', `anomalous-tts-state ${item.state === 'error' ? 'is-error' : 'is-missing'}`,
+        item.state === 'error' ? t('ttsPretrainedError') : t(item.required ? 'ttsPretrainedMissing' : 'ttsPretrainedOptional'));
+    if (item.error) state.title = item.error;
+    return state;
+}
+
+function renderPretrained(status, local, onChanged) {
+    const group = el('div', 'anomalous-tts-setup-group');
+    group.append(el('div', 'anomalous-tts-section', t('ttsSetupPretrained')));
+    const missing = status.pretrained.filter(item => item.state === 'missing' || item.state === 'error');
+    for (const item of status.pretrained) {
+        const row = el('div', 'anomalous-tts-setup-row');
+        const label = el('span', 'anomalous-tts-setup-path', item.label);
+        label.title = item.path || item.error || '';
+        row.append(label, el('span', 'anomalous-tts-kind', t(`ttsNeededFor_${item.needed_for}`)), pretrainedState(item));
+        if (item.state === 'missing' || item.state === 'error') {
+            const get = button('anomalous-tts-add', t('ttsPretrainedDownload', { size: formatSize(item.size) }),
+                () => act(get, () => startPretrainedDownload([item.id]), onChanged));
+            get.disabled = !local;
+            row.append(get);
+        }
+        group.append(row);
+    }
+    const actions = el('div', 'anomalous-tts-setup-actions');
+    if (missing.length > 1) {
+        const total = missing.reduce((sum, item) => sum + item.size, 0);
+        const all = button('anomalous-tts-add', t('ttsPretrainedDownloadAll', { size: formatSize(total) }),
+            () => act(all, () => startPretrainedDownload(missing.map(item => item.id)), onChanged));
+        all.disabled = !local;
+        actions.append(all);
+    }
+    const useSource = button('anomalous-tts-add', t('ttsPretrainedUsePackage'), () => act(useSource, async () => {
+        const path = await pickServerPath({ mode: 'folder', title: t('ttsPretrainedUsePackage'), hint: t('ttsPretrainedPackageHint') });
+        if (!path) return false;
+        await changePretrainedSource(path);
+    }, onChanged));
+    useSource.disabled = !local;
+    actions.append(useSource);
+    group.append(actions);
+    for (const source of status.pretrained_sources || []) {
+        const row = el('div', 'anomalous-tts-setup-row');
+        const path = el('span', 'anomalous-tts-setup-path', source);
+        path.title = source;
+        const remove = button('anomalous-tts-remove', '×', () => act(remove, () => changePretrainedSource(source, true), onChanged), t('ttsLibraryRemove'));
+        remove.disabled = !local;
+        row.append(path, el('span', 'anomalous-tts-kind', t('ttsPretrainedSource')), remove);
+        group.append(row);
+    }
+    group.append(el('div', 'anomalous-tts-hint', t('ttsPretrainedHint')));
+    return group;
+}
+
+function renderPackages(status) {
+    const missing = Object.entries(status.dependencies || {}).filter(([, dep]) => !dep.ok);
+    if (!missing.length) return null;
+    const group = el('div', 'anomalous-tts-setup-group');
+    group.append(el('div', 'anomalous-tts-section', t('ttsSetupPackages')));
+    for (const [lang, dep] of missing) {
+        const row = el('div', 'anomalous-tts-setup-row is-package');
+        const code = el('code', 'anomalous-tts-setup-command', dep.command);
+        const copy = button('anomalous-tts-add', t('ttsCopyCommand'), async () => {
+            try { await navigator.clipboard.writeText(dep.command); copy.textContent = t('ttsCopied'); } catch (_) { /* select by hand */ }
+        });
+        row.append(el('span', 'anomalous-tts-setup-meta', t('ttsPackagesFor', { lang: t(`ttsNeededFor_${lang}`), packages: dep.missing.join(', ') })), code, copy);
+        group.append(row);
+    }
+    group.append(el('div', 'anomalous-tts-hint', t('ttsPackagesHint')));
+    return group;
+}
+
+/**
+ * The card for one status payload. `onChanged()` redraws the studio (libraries and
+ * sources change the character list); `onImport()` opens the import form.
+ */
+export function renderTtsSetup(status, { onChanged, onImport }) {
+    const summary = setupSummary(status);
+    const local = status.local !== false;
+    const card = el('section', `anomalous-tts-setup${summary.ready ? ' is-ready' : ''}`);
+    const stored = storedOpen();
+    let open = stored === null ? !summary.ready : stored === '1';
+
+    const head = el('div', 'anomalous-tts-setup-head');
+    const toggle = button('anomalous-tts-setup-toggle', '', () => {
+        open = !open;
+        storeOpen(open);
+        card.classList.toggle('is-open', open);
+        toggle.setAttribute('aria-expanded', String(open));
+    });
+    toggle.setAttribute('aria-expanded', String(open));
+    toggle.append(el('span', 'anomalous-tts-setup-dot'), el('span', 'anomalous-tts-setup-title', t('ttsSetupTitle')),
+        el('span', 'anomalous-tts-setup-summary', summaryText(status, summary)), el('span', 'anomalous-tts-setup-chevron', '▾'));
+    const importBtn = button('anomalous-voice-modal-submit anomalous-tts-setup-import', t('ttsImportOpen'), onImport);
+    importBtn.disabled = !local;
+    if (!local) importBtn.title = t('ttsSetupRemote');
+    head.append(toggle, importBtn);
+
+    const body = el('div', 'anomalous-tts-setup-body');
+    if (!local) body.append(el('div', 'anomalous-tts-setup-notice', t('ttsSetupRemote')));
+    body.append(renderLibraries(status, local, onChanged), renderPretrained(status, local, onChanged));
+    const packages = renderPackages(status);
+    if (packages) body.append(packages);
+
+    card.classList.toggle('is-open', open);
+    card.append(head, body);
+    if (summary.downloading) pollWhileDownloading(card, { onChanged, onImport });
+    return card;
+}
+
+function pollWhileDownloading(card, options) {
+    setTimeout(async () => {
+        if (!card.isConnected) return;
+        try {
+            const status = await loadGptSovitsStatus({ force: true });
+            if (card.isConnected && status) card.replaceWith(renderTtsSetup(status, options));
+        } catch (_) {
+            if (card.isConnected) pollWhileDownloading(card, options);
+        }
+    }, POLL_MS);
+}

web/styles.css

@@ -10,5 +10,6 @@
 @import url("./styles/09-prompt-drawers.css?v=20260918-model-sources-1");
 @import url("./styles/10-model-sources.css?v=20260918-model-sources-1");
 @import url("./styles/11-audio-studio.css?v=20260925-tts-cards-1");
+@import url("./styles/12-tts-setup.css?v=20260925-tts-setup-1");
 @import url("./styles/13-script-director.css?v=20260923");
 @import url("./styles/14-version-manager.css");

web/styles/12-tts-setup.css

@@ -0,0 +1,398 @@
+/* GPT-SoVITS setup card, import form, server path picker and studio file drop.
+   Modal shell, fields, chips and buttons come from 11-audio-studio.css. */
+
+/* ---------- Setup card ---------- */
+.anomalous-tts-setup {
+    margin: 0 0 12px;
+    border: 1px solid rgba(245, 158, 11, 0.35);
+    border-radius: 12px;
+    background: rgba(245, 158, 11, 0.05);
+    color: #cbd5e1;
+    font-size: 12px;
+}
+
+.anomalous-tts-setup.is-ready {
+    border-color: rgba(52, 211, 153, 0.25);
+    background: rgba(52, 211, 153, 0.04);
+}
+
+.anomalous-tts-setup-head {
+    display: flex;
+    align-items: center;
+    gap: 10px;
+    padding: 8px 10px 8px 12px;
+}
+
+.anomalous-tts-setup-toggle {
+    display: flex;
+    flex: 1;
+    min-width: 0;
+    align-items: center;
+    gap: 8px;
+    padding: 2px 0;
+    border: none;
+    background: none;
+    color: inherit;
+    font: inherit;
+    text-align: left;
+    cursor: pointer;
+}
+
+.anomalous-tts-setup-dot {
+    flex-shrink: 0;
+    width: 8px;
+    height: 8px;
+    border-radius: 50%;
+    background: #f59e0b;
+}
+
+.anomalous-tts-setup.is-ready .anomalous-tts-setup-dot {
+    background: #34d399;
+}
+
+.anomalous-tts-setup-title {
+    flex-shrink: 0;
+    color: #f1f5f9;
+    font-weight: 700;
+}
+
+.anomalous-tts-setup-summary {
+    min-width: 0;
+    overflow: hidden;
+    color: #94a3b8;
+    text-overflow: ellipsis;
+    white-space: nowrap;
+}
+
+.anomalous-tts-setup-chevron {
+    margin-left: auto;
+    color: #64748b;
+    transition: transform 0.15s ease;
+}
+
+.anomalous-tts-setup.is-open .anomalous-tts-setup-chevron {
+    transform: rotate(180deg);
+}
+
+.anomalous-voice-modal-submit.anomalous-tts-setup-import {
+    flex-shrink: 0;
+    padding: 5px 14px;
+    font-size: 12px;
+}
+
+.anomalous-tts-setup-body {
+    display: none;
+    flex-direction: column;
+    gap: 14px;
+    padding: 4px 12px 12px;
+}
+
+.anomalous-tts-setup.is-open .anomalous-tts-setup-body {
+    display: flex;
+}
+
+.anomalous-tts-setup-group {
+    display: flex;
+    flex-direction: column;
+    gap: 6px;
+}
+
+.anomalous-tts-setup-row {
+    display: flex;
+    align-items: center;
+    gap: 8px;
+    min-height: 26px;
+}
+
+.anomalous-tts-setup-row.is-package {
+    flex-wrap: wrap;
+}
+
+.anomalous-tts-setup-path {
+    flex: 1;
+    min-width: 0;
+    overflow: hidden;
+    color: #e2e8f0;
+    text-overflow: ellipsis;
+    white-space: nowrap;
+}
+
+.anomalous-tts-setup-meta {
+    flex-shrink: 0;
+    color: #94a3b8;
+}
+
+.anomalous-tts-setup-command {
+    flex: 1 1 320px;
+    min-width: 0;
+    padding: 4px 8px;
+    border-radius: 6px;
+    background: rgba(0, 0, 0, 0.3);
+    color: #e2e8f0;
+    font: 11px/1.5 ui-monospace, SFMono-Regular, Menlo, Consolas, monospace;
+    overflow-wrap: anywhere;
+}
+
+.anomalous-tts-setup-actions {
+    display: flex;
+    flex-wrap: wrap;
+    gap: 6px;
+}
+
+.anomalous-tts-setup-notice {
+    padding: 8px 10px;
+    border-radius: 8px;
+    background: rgba(248, 113, 113, 0.08);
+    color: #fca5a5;
+}
+
+.anomalous-tts-kind {
+    flex-shrink: 0;
+    padding: 1px 6px;
+    border-radius: 4px;
+    background: rgba(129, 140, 248, 0.12);
+    color: #a5b4fc;
+    font-size: 10.5px;
+    white-space: nowrap;
+}
+
+.anomalous-tts-state {
+    flex-shrink: 0;
+    font-size: 11px;
+    white-space: nowrap;
+}
+
+.anomalous-tts-state.is-ok { color: #34d399; }
+.anomalous-tts-state.is-busy { color: #38bdf8; }
+.anomalous-tts-state.is-missing { color: #f59e0b; }
+.anomalous-tts-state.is-error { color: #f87171; cursor: help; }
+
+.anomalous-tts-add:disabled,
+.anomalous-tts-remove:disabled,
+.anomalous-voice-modal-submit.anomalous-tts-setup-import:disabled {
+    opacity: 0.5;
+    cursor: not-allowed;
+}
+
+/* ---------- Import form ---------- */
+.anomalous-tts-import {
+    width: 720px;
+}
+
+.anomalous-tts-import-target {
+    display: grid;
+    grid-template-columns: minmax(0, 1fr) minmax(0, 1.3fr);
+    gap: 10px;
+}
+
+.anomalous-tts-import-target:empty {
+    display: none;
+}
+
+.anomalous-tts-import-drop {
+    cursor: default;
+}
+
+.anomalous-tts-import-list {
+    display: flex;
+    flex-direction: column;
+    gap: 6px;
+}
+
+.anomalous-tts-import-row {
+    display: flex;
+    flex-direction: column;
+    gap: 6px;
+    padding: 7px 8px;
+    border: 1px solid rgba(255, 255, 255, 0.08);
+    border-radius: 8px;
+    background: rgba(0, 0, 0, 0.2);
+}
+
+.anomalous-tts-import-line,
+.anomalous-tts-import-audio {
+    display: flex;
+    align-items: center;
+    gap: 8px;
+    min-width: 0;
+}
+
+/* The play button's own display rule would override [hidden] (local paths cannot be played). */
+.anomalous-tts-import-audio [hidden] {
+    display: none;
+}
+
+.anomalous-tts-import-audio input {
+    min-width: 0;
+    padding: 5px 8px;
+    border: 1px solid rgba(255, 255, 255, 0.1);
+    border-radius: 6px;
+    background: rgba(0, 0, 0, 0.25);
+    color: #e2e8f0;
+    font-size: 12px;
+}
+
+.anomalous-tts-import-audio .anomalous-tts-name-input {
+    flex: 0 0 120px;
+}
+
+.anomalous-tts-import-audio .anomalous-tts-text-input {
+    flex: 1;
+}
+
+.anomalous-tts-import-audio .anomalous-tts-name-input:disabled {
+    opacity: 0.4;
+}
+
+.anomalous-tts-import-state {
+    flex-shrink: 0;
+    margin-left: auto;
+    color: #94a3b8;
+    font-size: 11px;
+}
+
+.anomalous-tts-import-state.is-error {
+    max-width: 260px;
+    overflow: hidden;
+    color: #f87171;
+    text-overflow: ellipsis;
+    white-space: nowrap;
+}
+
+.anomalous-tts-import-language {
+    max-width: 260px;
+}
+
+.anomalous-tts-import-problems {
+    display: flex;
+    flex-direction: column;
+    gap: 4px;
+}
+
+.anomalous-tts-import-problem {
+    padding: 5px 8px;
+    border-radius: 6px;
+    background: rgba(245, 158, 11, 0.08);
+    color: #fbbf24;
+    font-size: 11.5px;
+}
+
+.anomalous-tts-import-problem.is-error {
+    background: rgba(248, 113, 113, 0.08);
+    color: #f87171;
+}
+
+@media (max-width: 600px) {
+    .anomalous-tts-import-target {
+        grid-template-columns: 1fr;
+    }
+
+    .anomalous-tts-import-audio {
+        flex-wrap: wrap;
+    }
+}
+
+/* ---------- Server path picker ---------- */
+.anomalous-tts-picker {
+    width: 620px;
+}
+
+.anomalous-tts-picker-bar {
+    display: flex;
+    gap: 6px;
+}
+
+.anomalous-tts-picker-path {
+    flex: 1;
+    min-width: 0;
+    padding: 6px 8px;
+    border: 1px solid rgba(255, 255, 255, 0.1);
+    border-radius: 6px;
+    background: rgba(0, 0, 0, 0.25);
+    color: #e2e8f0;
+    font-size: 12px;
+}
+
+.anomalous-tts-picker-list {
+    display: flex;
+    flex-direction: column;
+    height: 320px;
+    overflow-y: auto;
+    border: 1px solid rgba(255, 255, 255, 0.08);
+    border-radius: 8px;
+    background: rgba(0, 0, 0, 0.2);
+}
+
+.anomalous-tts-picker-item {
+    display: flex;
+    flex-shrink: 0;
+    align-items: center;
+    gap: 8px;
+    padding: 6px 10px;
+    border: none;
+    border-bottom: 1px solid rgba(255, 255, 255, 0.04);
+    background: none;
+    color: #e2e8f0;
+    font: inherit;
+    font-size: 12px;
+    text-align: left;
+    cursor: pointer;
+}
+
+.anomalous-tts-picker-item:hover,
+.anomalous-tts-picker-item:focus-visible {
+    background: rgba(129, 140, 248, 0.1);
+}
+
+.anomalous-tts-picker-icon {
+    display: inline-flex;
+    color: #fbbf24;
+}
+
+.anomalous-tts-picker-name {
+    flex: 1;
+    min-width: 0;
+    overflow: hidden;
+    text-overflow: ellipsis;
+    white-space: nowrap;
+}
+
+.anomalous-tts-picker-meta {
+    flex-shrink: 0;
+    color: #64748b;
+    font-size: 11px;
+}
+
+.anomalous-tts-hint.is-error {
+    color: #f87171;
+}
+
+/* ---------- Dropping files on the studio ---------- */
+.anomalous-audio-studio-wrapper.is-file-dragover {
+    position: relative;
+    outline: 2px dashed rgba(129, 140, 248, 0.6);
+    outline-offset: -2px;
+    border-radius: 12px;
+}
+
+.anomalous-audio-studio-wrapper.is-file-dragover::after {
+    content: attr(data-drop-hint);
+    position: absolute;
+    top: 8px;
+    left: 50%;
+    z-index: 2;
+    padding: 6px 14px;
+    border-radius: 999px;
+    background: #4f46e5;
+    color: #fff;
+    font-size: 12px;
+    font-weight: 600;
+    white-space: nowrap;
+    transform: translateX(-50%);
+    pointer-events: none;
+}
+
+.anomalous-character-voice-card.is-file-target {
+    outline: 2px solid #818cf8;
+    outline-offset: 1px;
+}