Commit Graph

  • 1b37f259ee Two stages: test that a planted massive channel is found and steered in both stages Claude 2026-10-29 12:58:45 +00:00
  • dd80334d11 STAS: the Inactive notice names the strongest channel it found Claude 2026-10-29 12:57:12 +00:00
  • bd30a9524e Two stages: wrappers behind it wrap the whole run once; Noise merging, Seed search, Camera move, Shot memory follow the cut Claude 2026-10-29 12:52:25 +00:00
  • 8202caef5f Resolution hint: name the choice each example belongs to Claude 2026-10-29 11:55:29 +00:00
  • 5df050a7f2 Project settings: Resolution is the Short side / Long side / Pixel count; H3 sizes with FunUIAspectSize, old projects keep Pixel count Claude 2026-10-29 11:47:13 +00:00
  • c5eeceb42d Two stages: the whole-schedule copy goes to the stage schedule's device (CPU copy broke SLA and Decisiveness on the GPU) Claude 2026-10-29 11:25:26 +00:00
  • e439214b76 MXFP8: drop the claim that a forced compute dtype costs the speed (measured: same step times) Claude 2026-10-29 11:21:09 +00:00
  • f854f64c28 Two stages: both stages see the whole schedule (Decisiveness, SLA ramp, step windows count 8 steps, not 4 + 4) Claude 2026-10-29 11:16:28 +00:00
  • 9278d17f8b MXFP8: also allowed with weight dtype bf16 (the usual setting for a bf16 file) Claude 2026-10-29 11:10:57 +00:00
  • 51acdb1c12 Diffusion Model Loader: MXFP8 result goes to the log, not only the status output Claude 2026-10-29 11:07:48 +00:00
  • 373f0ea1ab Two stages: register as a modifier (it was never installed), test that every install() is provided Claude 2026-10-29 11:04:16 +00:00
  • 3837d585e0 VAE Loader: optional decode-only VAE (Sol-H3-VAE, X2 Detail) while the main VAE encodes Claude 2026-10-29 10:59:06 +00:00
  • a80e58490c Two stages as a sampling setting (a model wrapper, any sampler) instead of a preset Claude 2026-10-29 10:05:19 +00:00
  • 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 Claude 2026-10-29 09:55:37 +00:00
  • b108d4f4f5 H3 loader: MXFP8 option for the middle blocks (Sol-H3 recipe), credit in README Claude 2026-10-29 09:44:49 +00:00
  • 0f7bbedc60 5.1.0: note for LTX users on picking images and switching presets Claude 2026-10-27 23:24:26 +00:00
  • 6459b144f6 LTX: older Image to Video pipelines get the text to video fallback on load; end-to-end test of the no-input chain Claude 2026-10-27 12:23:29 +00:00
  • 3549907a6a 5.1.0: review fixes (interrupt in Noise merging, identity bank probe passes, explorer ratings per model family, NaN guide position), version and changelog Claude 2026-10-27 12:17:33 +00:00
  • 61dc98aeb0 Hide unproven enhancements on LTX: a model can hold modules back with hides:<id> traits Claude 2026-10-27 12:10:39 +00:00
  • 048dcd48cf Blocks settings adapt to the detected model: block count read from the file header, hint and clamp in the panel Claude 2026-10-27 10:56:02 +00:00
  • 92fe909a40 Attention explorer on LTX: depth groups and clip regions (prompt focus stays H3) Claude 2026-10-27 10:48:53 +00:00
  • eba7c63e9d Identity bank for LTX: BachVid K/V bank per Taste key with a KV-Lock schedule Claude 2026-10-27 10:41:37 +00:00
  • 8e22f17b00 LTX: Image to Video from References saved before the aspect source is upgraded on load Claude 2026-10-27 10:46:18 +00:00
  • 5d84c71056 LTX Image to Video: the timeline picture is the aspect source (new FunUI Aspect Size node), the project size is the budget Claude 2026-10-27 10:43:44 +00:00
  • a17484a9aa Freeze the Guide Frames node's inputs Claude 2026-10-27 10:37:54 +00:00
  • 617dbf2168 Guide frames for LTX (middle frame and memory of the clips before), REINS/Block repeat notes, README thanks JoyAI-Echo Claude 2026-10-27 10:32:30 +00:00
  • 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 Claude 2026-10-27 10:34:48 +00:00
  • 2f4a291b2e REINS and Block repeat work on LTX's (video, audio) blocks: new block_rows trait Claude 2026-10-27 10:26:44 +00:00
  • ee56a282d9 Randomizer (attention temperature) now works on LTX: new block_scope trait, LTX blocks counted Claude 2026-10-27 10:23:26 +00:00
  • 478d9d3af7 Temporal style auto picks a style per shot from its words; registry node name back to ComfyUI-FunPack (display name stays FunUI) Claude 2026-10-27 10:20:50 +00:00
  • cd0ef39cb1 LTX: drop the 'tested' log notice, a working model needs none Claude 2026-10-27 10:09:45 +00:00
  • 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 Claude 2026-10-27 09:44:54 +00:00
  • 7b9acb5fc1 Shot memory and Camera move: say a start video also counts as a non-empty start Claude 2026-10-27 09:43:55 +00:00
  • 3776c22364 Restoring a saved pipeline no longer lays in another preset's input slots (the identity loader landed in Video to Video, so a picture reference failed with 'identity: no such slot'); Video input card gets a width/height flip Claude 2026-10-27 09:32:34 +00:00
  • 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 Claude 2026-10-27 09:28:09 +00:00
  • 92ec6c27ce Project settings: a button that swaps width and height Claude 2026-10-27 09:18:02 +00:00
  • 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" Claude 2026-10-27 09:14:20 +00:00
  • 975ddf9027 Reapply "API client: a 200 that is not JSON is said plainly instead of failing as 'r is null'" Claude 2026-10-27 09:14:20 +00:00
  • 9d183daf4f Reapply "LTX image-to-video: the first reference picture is the start when none is picked" Claude 2026-10-27 09:14:20 +00:00
  • 06b467c4ed Reapply "LTX direct v2v: prompt path proven, start the schedule higher by default (0.909)" Claude 2026-10-27 09:14:20 +00:00
  • d277ebc76c Log: a console that cannot print a file name no longer fails the run that mentioned it Claude 2026-10-27 09:12:39 +00:00
  • 3a2d64c792 Load Media: decode only the frames a prepared video keeps, and only the first frame for a picture input Claude 2026-10-27 09:11:17 +00:00
  • f8955c2ef2 Keep ComfyUI's console on disk (Previous run in the Log window) and log memory at LoRA load and sampler start Claude 2026-10-27 09:03:30 +00:00
  • f2fcaefa86 Revert "LTX direct v2v: prompt path proven, start the schedule higher by default (0.909)" Claude 2026-10-27 08:53:24 +00:00
  • 5a1da0ce1b Revert "LTX image-to-video: the first reference picture is the start when none is picked" Claude 2026-10-27 08:52:11 +00:00
  • 4eb2ee2ee2 Revert "API client: a 200 that is not JSON is said plainly instead of failing as 'r is null'" Claude 2026-10-27 08:52:11 +00:00
  • 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" Claude 2026-10-27 08:52:11 +00:00
  • 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 Claude 2026-10-27 08:48:01 +00:00
  • eac7ecd582 API client: a 200 that is not JSON is said plainly instead of failing as 'r is null' Claude 2026-10-27 08:42:42 +00:00
  • e4111cf74a LTX image-to-video: the first reference picture is the start when none is picked Claude 2026-10-27 08:39:08 +00:00
  • 859ef99283 LTX direct v2v: prompt path proven, start the schedule higher by default (0.909) Claude 2026-10-27 08:32:10 +00:00
  • 9f8705dd63 Shot camera and SLA need H3-only traits: hidden on LTX (node fields can now require a trait) Claude 2026-10-27 08:29:22 +00:00
  • 11adad2fb4 LTX direct v2v: end-to-end test (real mp4, saved-pipeline upgrade, Load Media, encode, sampler on a real tiny LTXAV) Claude 2026-10-27 08:21:45 +00:00
  • 702aea0f08 LTX video-in pipelines saved with the first-frame wiring are upgraded on run Claude 2026-10-27 08:17:30 +00:00
  • 4ed0685dbc Video-in presets: wire the frames output statically (a wired input cannot be overridden per run) Claude 2026-10-27 08:13:18 +00:00
  • 1a5d9eab19 Video-to-video presets: the first video among the references is the start when none is picked Claude 2026-10-27 08:09:44 +00:00
  • a9f5c9f35d VAE loader: pass the file header to the VAE (audio VAEs need their config) Claude 2026-10-27 08:06:34 +00:00
  • 2a119b76f7 LTX presets: separate model, video VAE, audio VAE and text encoder loaders Claude 2026-10-27 08:01:58 +00:00
  • 2f5a494b0b LTX 2.3 direct Video to Video preset (encode the video, sample from part-way down the schedule) Claude 2026-10-27 07:47:24 +00:00
  • 36fd60c427 LTX 2.3 Video to Video preset after ComfyUI's IC-LoRA template Claude 2026-10-27 07:31:19 +00:00
  • 1e70272dde Video input: on LTX a video picked as the scene start feeds its frames, not just the first one Claude 2026-10-27 07:29:40 +00:00
  • a4eaabc20e Video input: prepare reference videos (size, crop, rate, skip, cut, every Nth, cap) per project or per scene Claude 2026-10-27 01:04:58 +00:00
  • 9d21739b11 Quick A/B: on the wheel as soon as it is ticked, not only after a run in this page session Claude 2026-10-26 19:42:14 +00:00
  • 677491a18d Attention explorer: prompt focus at 0.3 of the change size, new levers change less often at first Claude 2026-10-25 22:17:04 +00:00
  • 1b0adbd45c Quick A/B: ends with the previous clip and the new one, labelled, side by side Claude 2026-10-25 22:16:52 +00:00
  • 01b8a37dff Attention explorer: prompt weighting, prompt focus, per-prompt recipes, clip regions Claude 2026-10-25 22:10:19 +00:00
  • 5a98b2ed35 Picker wheel: Quick A/B (every enhancement off, last run's seed, generates the picked scene right away, then restores) Claude 2026-10-25 21:54:51 +00:00
  • 089248ea10 Attention explorer: dislikes no longer grow the change size, learned size capped at 0.35 Claude 2026-10-25 21:37:21 +00:00
  • f71407f98e Attention explorer: the log reports how many attention calls were actually changed, and warns if none Claude 2026-10-25 21:33:33 +00:00
  • 444c65a1de Attention explorer: the log line counts rated, liked and disliked runs Claude 2026-10-25 20:57:03 +00:00
  • 9699eeea61 Attention explorer: no cap on rated runs Claude 2026-10-25 20:52:51 +00:00
  • 06a8c7948a Attention explorer keeps up to 200 rated runs, not 64 Claude 2026-10-25 20:52:09 +00:00
  • 71b9a1f514 A rating re-runs the same seed for every learner, not only the Attention explorer Claude 2026-10-25 20:44:59 +00:00
  • 0fc20bbc3d Attention explorer: a rating re-runs the same seed instead of hitting ComfyUI's cache Claude 2026-10-25 20:43:03 +00:00
  • c8e9a00c86 Save the project when the page is left, and apply the media bin's saved view once the project opens Claude 2026-10-25 20:28:23 +00:00
  • 0599bcc676 Attention explorer: change size learns from ratings, log names exact blocks moved Claude 2026-10-25 20:20:38 +00:00
  • db46828c99 Attention explorer: per-depth-group attention temperature drawn from ratings Claude 2026-10-25 20:01:23 +00:00
  • 4b48674975 Camera enhancements wheel switch, stable Save video as name, media bin remembers view and sort Claude 2026-10-25 17:51:44 +00:00
  • db3be87e88 Noise merging: Second run mode, a whole second video from the second seed blended in at the named steps Claude 2026-10-25 17:46:51 +00:00
  • bc88ac57a7 Noise merging: Second pass new noise, how much of the step's own noise the second pass swaps for the second seed's Claude 2026-10-25 17:29:33 +00:00
  • e044819db0 Noise merging: optional merge of the starting noise (variance-preserving mix with a second seed) Claude 2026-10-25 17:25:06 +00:00
  • 8d5d6a72b7 Noise merging: Merge sound too option (on by default) Claude 2026-10-25 17:10:49 +00:00
  • b5656db6b8 Timing: the sampler line gives the time of every step, not just the first two Claude 2026-10-25 17:03:07 +00:00
  • bc7585cddb Noise merging: merge the sound too; test that run 2 and 3 on the same cached model still make the extra pass Claude 2026-10-25 17:00:58 +00:00
  • f44ebc6475 Noise merging: the log line says how long the second pass took Claude 2026-10-25 16:58:36 +00:00
  • db291b8c4b Noise merging: end-to-end test from the app's saved settings through the pipeline route to a 4-step lcm run on the real H3 model class Claude 2026-10-25 16:53:06 +00:00
  • 53c632d632 Noise merging: say it is on at the start of every run, and say so when no step matched Claude 2026-10-25 16:49:40 +00:00
  • 5ff02e65dd Noise merging: the log says how far apart the two answers were at each merged step Claude 2026-10-25 16:42:56 +00:00
  • dc3c580456 Noise merging: second seed is random each run by default (-1) Claude 2026-10-25 16:40:19 +00:00
  • 8a241cf9be Noise merging: answer chosen steps twice with a second seed and blend the two pictures Claude 2026-10-25 16:35:34 +00:00
  • cce20f136a Shot Camera: spaCy model failed to load with preshed 3.0.14 (OverflowError); pin preshed<3.0.14, repair on start, name the real cause in the log Claude 2026-10-25 16:20:09 +00:00
  • b903648950 Wheel: Clear references and Disable all guidance enhancements; Guidance switch in Engine settings; branch notes in the branch switcher Claude 2026-10-25 16:16:40 +00:00
  • 2ce29a0998 SLA rising sparsity: note the first real H3 result Claude 2026-10-25 16:16:02 +00:00
  • 5bc3ce3f95 SLA: test that the rising-share field is drawn only for the sla method Claude 2026-10-25 16:06:31 +00:00
  • 527d368c82 SLA: optional rising sparsity (sla_sparsity_start), GSRS idea; CPU test env script Claude 2026-10-25 16:05:33 +00:00
  • 29511f73e8 Registry: real FunUI icon (the wink on dark) Claude 2026-10-25 13:54:48 +00:00
  • 199c39ad98 PTD: the hop notice names the adapter each time it changes Claude 2026-10-25 13:48:17 +00:00
  • dc4e51563c 5.0.3: stale Decisiveness test, migration tests, changelog folded into the release Claude 2026-10-25 13:44:56 +00:00
  • 3fa2188a97 README: all screenshots retaken with the FunUI name and logo Claude 2026-10-25 13:39:33 +00:00
  • 2e445244df Test: self-removal refusal names the folder, whatever it is called Claude 2026-10-25 13:28:36 +00:00