nodes/olivv-cs-ComfyUI-FunPack
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
2,481 Commits 3 Branches 11 Tags
3dbcc8a4e8fada32db8f7be62357ef385ce18dab
Commit Graph
100 Commits
Author SHA1 Message Date
digital-garbageandClaude Sonnet 5.5 9b1cb52024 Post settings in Engine > Post: smoothing moved out of the node, plus auto contrast, auto brightness, colour strength 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-11-05 20:32:03 +02:00
digital-garbage d792d7f5e9 Matrix: LTX file, LoRA slots across stages, stale KJNodes skips removed 
Every modifier alone, in pairs and all together on tiny H3 and tiny LTX; the H3 preset build tests run again.
 2026-11-05 16:01:20 +02:00
Claude e2f14bd1a7 Continuity: H3 only for now, hint recommends R2V 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjYBYFyV8YG5YDfYPRb
 2026-11-02 13:37:55 +00:00
Claude 6fad1d4feb Continuity: carried frame takes the next free reference slot 
It was fed to the aspect-ratio picture slot only. Hooks may now return references, added to the scene's before wiring.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjYBYFyV8YG5YDfYPRb
 2026-11-02 13:27:13 +00:00
Claude 4c0192469a Presets can carry project settings; Fast & Simple R2V turns on Frame Generation (always, Balanced x2) and Super Resolution on demand (Denoise Medium) 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjWzZx5QzcqkheSuhsy
 2026-10-29 19:49:44 +00:00
Claude 15bd8b8480 H3 Fast & Simple R2V preset: MXFP8, ramped sparse attention, DMAD LoRA slot, 5-step Turbo with Two stages and polish, Attention explorer, camera moves, Sol decode-only VAE 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjWzZx5QzcqkheSuhsy
 2026-10-29 19:19:17 +00:00
Claude 5df050a7f2 Project settings: Resolution is the Short side / Long side / Pixel count; H3 sizes with FunUIAspectSize, old projects keep Pixel count 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjYcwzhZeJZAD6ujMjA
 2026-10-29 11:47:13 +00:00
Claude a80e58490c Two stages as a sampling setting (a model wrapper, any sampler) instead of a preset 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjW75aYB8UyVRZo6fCr
 2026-10-29 10:05:19 +00:00
Claude 791bb6ca48 H3: two-stage preset (first steps on a smaller picture), Sampler denoised output, Resample to an exact size without an upscaler file, Stage Size node 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjW75aYB8UyVRZo6fCr
 2026-10-29 09:55:37 +00:00
Claude 6459b144f6 LTX: older Image to Video pipelines get the text to video fallback on load; end-to-end test of the no-input chain 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjYZhHqk196LykrWy3T
 2026-10-27 12:23:29 +00:00
Claude 61dc98aeb0 Hide unproven enhancements on LTX: a model can hold modules back with hides:<id> traits 
Only Noise merging and the Attention explorer (and Taste) stay offered on LTX until the rest have been proven there.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjYZhHqk196LykrWy3T
 2026-10-27 12:10:39 +00:00
Claude 048dcd48cf Blocks settings adapt to the detected model: block count read from the file header, hint and clamp in the panel 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjYZhHqk196LykrWy3T
 2026-10-27 10:56:02 +00:00
Claude 8e22f17b00 LTX: Image to Video from References saved before the aspect source is upgraded on load 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 10:46:18 +00:00
Claude 5d84c71056 LTX Image to Video: the timeline picture is the aspect source (new FunUI Aspect Size node), the project size is the budget 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 10:43:45 +00:00
Claude 617dbf2168 Guide frames for LTX (middle frame and memory of the clips before), REINS/Block repeat notes, README thanks JoyAI-Echo 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjYZhHqk196LykrWy3T
 2026-10-27 10:38:06 +00:00
Claude a63bedfd5f Tests: route tests get stand-in model files so a machine without models still sees a deleted file; optional-slot list follows the two LTX presets 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 10:34:49 +00:00
Claude 2f4a291b2e REINS and Block repeat work on LTX's (video, audio) blocks: new block_rows trait 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjYZhHqk196LykrWy3T
 2026-10-27 10:26:44 +00:00
Claude ee56a282d9 Randomizer (attention temperature) now works on LTX: new block_scope trait, LTX blocks counted 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjYZhHqk196LykrWy3T
 2026-10-27 10:23:26 +00:00
Claude cd0ef39cb1 LTX: drop the 'tested' log notice, a working model needs none 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Tw9fjYZhHqk196LykrWy3T
 2026-10-27 10:09:45 +00:00
Claude c70c74f01f LTX: two tested presets (Image to Video from References, Video to Video direct) with text to video as the fallback in both; docs say LTX 2.3 is tested 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 09:45:04 +00:00
Claude 29f7860532 LTX: Image to Video preset that starts from the first reference; a reference feeding a missing node is left unused instead of refusing the run 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 09:28:09 +00:00
Claude d4641450c8 Reapply "Direct v2v saved before the video-only flag keeps taking its video from references; render polls survive one unanswered reply; LTX project HTTP test" 
This reverts commit 5a2f9c5999.
 2026-10-27 09:14:20 +00:00
Claude 9d183daf4f Reapply "LTX image-to-video: the first reference picture is the start when none is picked" 
This reverts commit 5a1da0ce1b.
 2026-10-27 09:14:20 +00:00
Claude 06b467c4ed Reapply "LTX direct v2v: prompt path proven, start the schedule higher by default (0.909)" 
This reverts commit f2fcaefa86.
 2026-10-27 09:14:20 +00:00
Claude f2fcaefa86 Revert "LTX direct v2v: prompt path proven, start the schedule higher by default (0.909)" 
This reverts commit 859ef99283.
 2026-10-27 08:53:24 +00:00
Claude 5a1da0ce1b Revert "LTX image-to-video: the first reference picture is the start when none is picked" 
This reverts commit e4111cf74a.
 2026-10-27 08:52:11 +00:00
Claude 5a2f9c5999 Revert "Direct v2v saved before the video-only flag keeps taking its video from references; render polls survive one unanswered reply; LTX project HTTP test" 
This reverts commit ed4de1e8af.
 2026-10-27 08:52:11 +00:00
Claude ed4de1e8af Direct v2v saved before the video-only flag keeps taking its video from references; render polls survive one unanswered reply; LTX project HTTP test 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 08:48:01 +00:00
Claude e4111cf74a LTX image-to-video: the first reference picture is the start when none is picked 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 08:39:08 +00:00
Claude 859ef99283 LTX direct v2v: prompt path proven, start the schedule higher by default (0.909) 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 08:32:10 +00:00
Claude 9f8705dd63 Shot camera and SLA need H3-only traits: hidden on LTX (node fields can now require a trait) 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 08:29:22 +00:00
Claude 11adad2fb4 LTX direct v2v: end-to-end test (real mp4, saved-pipeline upgrade, Load Media, encode, sampler on a real tiny LTXAV) 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 08:21:45 +00:00
Claude 702aea0f08 LTX video-in pipelines saved with the first-frame wiring are upgraded on run 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 08:17:30 +00:00
Claude 4ed0685dbc Video-in presets: wire the frames output statically (a wired input cannot be overridden per run) 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 08:13:18 +00:00
Claude 2a119b76f7 LTX presets: separate model, video VAE, audio VAE and text encoder loaders 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 08:01:58 +00:00
Claude 2f5a494b0b LTX 2.3 direct Video to Video preset (encode the video, sample from part-way down the schedule) 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 07:47:24 +00:00
Claude 36fd60c427 LTX 2.3 Video to Video preset after ComfyUI's IC-LoRA template 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 07:31:19 +00:00
Claude 1e70272dde Video input: on LTX a video picked as the scene start feeds its frames, not just the first one 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BzJLJWVu7RZnbRqm3nYAnX
 2026-10-27 07:29:40 +00:00
Claude 3adebb731f Rename FunPack to FunUI (5.0.3 Classy Chopper): /fun endpoint, wink logo, load-time migration of old names 
Old project/workflow names and browser settings migrate on load; /funpack redirects to /fun.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016acRss69eH1235eRdFTZHY
 2026-10-25 13:11:03 +00:00
Claude 49d69aa394 test: a picked H3 file keeps Shadow negative and Video detail on screen 
Fake safetensors header through the probe and the real manifest; fails on the
code before 9b4b559, passes after.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011av2BXzE4UT5HfXpnhTNJQ
 2026-10-22 08:37:09 +00:00
Claude 9b4b559d19 fix: H3 file probe now shows Shadow negative and Video detail 
The pre-load probe left out adaln_modalities, so picking an H3 file hid every
setting that needs it. Run time still checks the loaded model.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011av2BXzE4UT5HfXpnhTNJQ
 2026-10-22 08:17:55 +00:00
Claude 329781499c feat: H3 preset has 15 reference slots (9 pictures, 3 videos, 3 sounds); saved projects get the new ones on open 
Pictures are numbered among pictures, so any mix fits R2V's own inputs.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CHWxqHtxwWrq4kEyFTaHds
 2026-10-21 02:54:22 +00:00
Claude a36af5f3a1 fix: a video or audio reference is routed from the inputs the model node offers, not from preset data 
Works for any H3 Reference to Video pipeline, edited or not. Drops the preset routing data and the refresh that carried it.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CHWxqHtxwWrq4kEyFTaHds
 2026-10-21 02:43:47 +00:00
Claude bf62683f5a feat: Load Media takes a picture, video or sound and the app wires the output that fits 
A video or audio reference now uses an ordinary reference slot, so old projects need no new slots; saved media slots take the preset's current roles.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CHWxqHtxwWrq4kEyFTaHds
 2026-10-21 02:22:23 +00:00
Claude 129cc10d0f fix: video and audio references go to R2V's own video/audio inputs; aspect ratio source accepts a video 
New Load Media (video) and (audio) nodes; picture loader keeps first frame for the aspect source only.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CHWxqHtxwWrq4kEyFTaHds
 2026-10-21 02:01:01 +00:00
Claude 919097c732 LoRA training: our own loop for H3, and a Log that survives a crash 
The H3 run trains with a loop of our own (LoRA on the attention and feed-forward projections of the
50 blocks, per-block checkpoint, uniform noise levels, musubi-tuner's approach) while ComfyUI carries
the batches. Every Log record is flushed to disk as it is said and reloaded on start; the run
writes each stage and step to it. ComfyUI's own offloading toggle is ignored by the new loop.
UNTESTED on the real weights.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MpnF7GwNGMbngU11wjym7P
 2026-10-17 19:57:06 +00:00
Claude a13f037f72 H3 training: modulate blocks out of place so the backward pass stops keeping every block's recompute alive 
Measured on a tiny H3: backward growth was ~75 MB per block, now flat (61-88 MB at 2, 8 and 16 blocks).
The in-place mul_/addcmul_ on slices of the stream in the block helpers were the cause.
UNTESTED on the real weights.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MpnF7GwNGMbngU11wjym7P
 2026-10-17 19:47:34 +00:00
Claude d0d572c4ec feat: block swap option for LoRA training and a per-block memory log in the backward pass 
Block swap keeps the frozen weights of the last N transformer blocks in system
memory and lets ComfyUI's weight casting fetch them one layer at a time. The idea
and the block count as the one knob follow musubi-tuner's block swap (kohya-ss,
Apache License 2.0); this is an independent implementation.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MpnF7GwNGMbngU11wjym7P
 2026-10-17 19:30:50 +00:00
Claude a070ce9400 Training reads every clip for as many frames as the shortest gives, and allows up to 1025 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MpnF7GwNGMbngU11wjym7P
 2026-10-17 18:30:50 +00:00
Claude bb9ec64a46 H3 training checkpoints one block at a time and the text refiner on its own 
Depth 1 made the whole model a single checkpoint, so nothing was saved.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MpnF7GwNGMbngU11wjym7P
 2026-10-17 18:14:55 +00:00