nodes/olivv-cs-ComfyUI-FunPack
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
1,778 Commits 3 Branches 11 Tags
768838c3d7bbddacd7adade18f7276dfe27fda7f
Commit Graph
451 Commits
Author SHA1 Message Date
digital-garbageandClaude Sonnet 5.5 768838c3d7 feat: camera moves, styles, targets and shot splits learn from ratings on their own 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-10-01 23:07:52 +03:00
digital-garbage fa232c5e76 feat: views work on a lone shot; per-shot Forget for saved choices, saved immediately 2026-09-30 21:03:52 +03:00
digital-garbage 80808442eb feat: the shot-views log says why each shot got no view 2026-09-30 20:58:37 +03:00
digital-garbage cfc349bf85 fix: body actions 'from behind/above/below' are not stated camera views 2026-09-30 20:52:58 +03:00
digital-garbage 55d6820296 feat: read and forget learned camera focus / view memory in Settings 2026-09-30 20:46:59 +03:00
digital-garbage b734e5a2e6 feat: Reactive focus shows shot previews and takes several views / ordered targets per shot 2026-09-30 20:43:27 +03:00
digital-garbage 13e2001f47 fix: Reactive focus review crashed on projects with $variables; failures now show a notice 2026-09-30 20:30:51 +03:00
digital-garbageandClaude Sonnet 5.5 45535f4d91 feat: shot views learn from ratings, respect what a shot shows, and get a Reactive focus step 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 20:19:56 +03:00
digital-garbageandClaude Sonnet 5.5 5d419c97d9 feat: reactive focus backend: per-shot focus options, user choices and learning 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 20:10:19 +03:00
digital-garbageandClaude Sonnet 5.5 e91fc71150 feat: camera moves include static holds on one object; X to Y fires less often 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 20:01:04 +03:00
digital-garbageandClaude Sonnet 5.5 8bbed2b1f8 refactor: the recurring-word memory is its own file, not tied to a refinement key or the host pack 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 19:15:18 +03:00
digital-garbageandClaude Sonnet 5.5 b5fb00c6a5 feat: habit-weighted chance for camera targets, in-shot prominence, conjoined nouns share a role 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 19:11:19 +03:00
digital-garbageandClaude Sonnet 5.5 459daebbc2 feat: camera targets favour words the user's prompts keep dwelling on (per key) 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 19:08:22 +03:00
digital-garbageandClaude Sonnet 5.5 1901f03fa5 fix: camera targets are whole phrases (base OF what, skin WHOSE) or not used 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 18:23:37 +03:00
digital-garbageandClaude Sonnet 5.5 2abd7fc30f fix: the rewriter's own cut openers and views never become camera targets 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 18:17:04 +03:00
digital-garbageandClaude Sonnet 5.5 efc5757950 fix: 'pull out' / 'push in' as a body action no longer counts as an existing camera move 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 18:15:20 +03:00
digital-garbageandClaude Sonnet 5.5 ccaf04bd70 fix: the closing camera move says pulls back, never pulls out 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 18:12:58 +03:00
digital-garbageandClaude Sonnet 5.5 c3c3f0fe96 fix: cut times split the scene evenly, no shot under 2 s 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 18:07:34 +03:00
digital-garbageandClaude Sonnet 5.5 389fbba275 fix: an unlinked H3 latent length follows the project's frames; export shows it 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 18:05:29 +03:00
digital-garbageandClaude Sonnet 5.5 b036a533aa fix: a POV view is not a camera move, so a shot given a view still gets its move 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 18:00:00 +03:00
digital-garbageandClaude Sonnet 5.5 8e4989ec09 fix: settings export hides stale loader inputs and SLA settings while SLA is off 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 17:55:18 +03:00
digital-garbageandClaude Sonnet 5.5 c5ae2d8163 fix: shots are only cut between two shortcuts, never inside one 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 17:46:34 +03:00
digital-garbageandClaude Sonnet 5.5 8e0c320af6 feat: shot cut times and shot views for H3 prompts, Editor toggles 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 17:31:48 +03:00
digital-garbageandClaude Sonnet 5.5 f87edd0895 feat: Composer shows the final prompt of every run, enhancer on or off 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 17:23:24 +03:00
digital-garbageandClaude Sonnet 5.5 4a2235591f fix: a noun shared by two of three shots is still a camera target 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 17:17:33 +03:00
digital-garbageandClaude Sonnet 5.5 62c481f77e fix: body movement and looking at the camera no longer count as an existing camera move 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 17:15:04 +03:00
digital-garbageandClaude Sonnet 5.5 4b182524e6 feat: camera-move wording follows H3's own prompt guide; a shot opening with a cut still gets a move 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 17:12:23 +03:00
digital-garbageandClaude Sonnet 5.5 0fc080b057 feat: camera moves are randomised per shot (chance, 1-3 moves, X to Y or not) and declare the camera 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 17:09:03 +03:00
digital-garbageandClaude Sonnet 5.5 4c30deb6a2 feat: camera moves travel from the first topic of a shot to the last 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 17:05:23 +03:00
digital-garbageandClaude Sonnet 5.5 f57910da2d feat: rule-based camera moves for [Shot N] blocks, Editor toggle 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 16:47:07 +03:00
digital-garbageandClaude Sonnet 5.5 a01ffbd154 feat: image-vs-composition dislikes and effect-weighted ratings for the rated dials 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 16:18:30 +03:00
digital-garbageandClaude Sonnet 5.5 942202c963 fix: camera move runs on GPU, stands down for multi-call samplers and says so, H3-only in the Editor 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 16:02:16 +03:00
digital-garbageandClaude Sonnet 5.5 6d606b881c feat: camera move works on the running latent from a chosen step (starting-noise version removed) 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 15:53:53 +03:00
digital-garbageandClaude Sonnet 5.5 0bdf070fba fix: camera noise declares the FreeNoise collision, matches the probe path, clamps pan; states the zoom-in cost 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 15:35:02 +03:00
digital-garbageandClaude Sonnet 5.5 aec5ac4a50 feat: camera move (pan/zoom/focus) written into the starting noise 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 15:30:49 +03:00
digital-garbageandClaude Sonnet 5.5 aa352de959 fix: decisiveness declares a schedule it cannot act on and stops learning there; new features' logs colour by state 
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
 2026-09-30 15:18:54 +03:00
digital-garbageandClaude Opus 5.5 81d8a9b094 feat: upscale finished renders with an upscale model (never / always / Upscale button) 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 22:30:37 +03:00
digital-garbageandClaude Opus 5.5 c072a088ec fix: input steer says when a 1-step schedule leaves it nothing to do 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 21:40:25 +03:00
digital-garbageandClaude Opus 5.5 915ba4a65b fix: input steer never hands a push to another call of the same step; size change is said 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 21:32:03 +03:00
digital-garbageandClaude Opus 5.5 9b53eecddb fix: H3 step wrappers steer the next step's input, never the output step 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 21:27:15 +03:00
digital-garbageandClaude Opus 5.5 c782a63fdb fix: shot memory only ever turns a reused layout down, never up (few-cell frames) 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 20:50:47 +03:00
digital-garbageandClaude Opus 5.5 454edace23 fix: shot memory skips frames too small for a layout (1-cell grid made the noise NaN) 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 20:46:00 +03:00
digital-garbageandClaude Opus 5.5 e87fde3c2d fix: shot memory keeps a normal-strength layout when a shot is reused with its own seed 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 20:37:50 +03:00
digital-garbageandClaude Opus 5.5 8bc7cb643e fix: late-branch guidance no longer leaks the model 
Its step wrapper held its own ModelPatcher, a cycle only a full gc frees;
that gc clears ComfyUI's weak link to the parent copy first, leaving it
tracking a dead model ("memory leak with model MiniMaxH3").

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 20:17:32 +03:00
digital-garbageandClaude Opus 5.5 106163d039 fix: STAS does not count late-branch weak calls; real-ComfyUI clone test 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 19:54:17 +03:00
digital-garbageandClaude Opus 5.5 aaa1c75240 fix: step wrappers find the picture/sound split on a ModelPatcher 
_get_latent_shapes only looked under a CFGGuider's conds; wrappers hold a
ModelPatcher, so on packed AV latents every video-only wrapper passed through
unchanged. Also: late-branch and STAS say when they applied nothing, and a
learned value is only rated when the feature actually applied.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 19:45:08 +03:00
digital-garbageandClaude Opus 5.5 ce2de22b2e feat: STAS for H3; late-branch guidance skips negative-prompt calls 
STAS (arXiv 2603.17825) sets massive-activation channels on frame 0 and
frame edges to alpha x peak at one block over the first 40% of steps.
Rating-learned dials now share rated_dial.py.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 19:35:29 +03:00
digital-garbageandClaude Opus 5.5 a9c2bd30c3 feat: late-branch guidance for H3, strength learned from ratings 
Skip guidance at CFG 1: the weak copy reuses blocks before the branch and
skips the branch block, so only the tail runs twice. Picture only; capture
hooks ignore the weak pass; compiler off while installed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 19:26:26 +03:00
digital-garbageandClaude Opus 5.5 63bd1f5ed6 feat: editable reference intro for the prompt enhancer; typo fixed 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 17:54:07 +03:00
digital-garbageandClaude Opus 5.5 2f3c9f5c1a feat: shot memory and decisiveness, both learned from ratings 
Shot memory reuses a liked gen's coarse starting noise (layout) with fresh
fine noise; reuse-vs-fresh, which shot and how strongly are learned.
Decisiveness is Temporal Score Rescaling (arXiv 2510.01184) on the picture
rows, k learned per key or set by hand.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-28 15:19:45 +03:00