nodes/havvk-ComfyUI_AIIA
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Commit Graph

  • 7d4937e8b8 fix(ditto): v1.9.56 - Restore Blink Indices but Attenuate Squint (0.2) + Boost Anti-Tilt (0.3) Hawk Lee 2026-01-23 00:35:40 +08:00
  • 78f1fdd87d fix(ditto): v1.9.55 - Revert to Legacy Blink Function (Original Code Logic) while keeping safe [11,13] mask Hawk Lee 2026-01-23 00:24:00 +08:00
  • 49cb9afcb9 feat(ditto): v1.9.54 - Add [Ditto Tilt Diag] logging needed for diagnosis Hawk Lee 2026-01-23 00:22:12 +08:00
  • c678a2b665 fix(ditto): v1.9.53 - Fix Blink Twitch by removing squint indices; Keep Anti-Tilt Hawk Lee 2026-01-23 00:19:17 +08:00
  • 688d92fe52 feat(ditto): v1.9.52 - Add specialized Auto-Centering for Roll Drift Hawk Lee 2026-01-23 00:11:45 +08:00
  • 7133328f54 feat(ditto): v1.9.51 - Stronger Gravity (0.2) & Louder Sway (2.5deg) Hawk Lee 2026-01-22 23:49:33 +08:00
  • a7a1ece946 feat(ditto): v1.9.50 - Fix Head Drift (Gravity Well) & Boost Idle Sway Hawk Lee 2026-01-22 23:36:24 +08:00
  • 403fcb4d26 feat(ditto): Positive-Only Mouth Breathing (v1.9.49) to fix Pursed Lips Hawk Lee 2026-01-22 23:14:13 +08:00
  • 9d44acce11 feat(ditto): Implement Mouth Micro-Motion (Breathing) during silence v1.9.48 Hawk Lee 2026-01-22 23:00:07 +08:00
  • 17acb10e47 fix(ditto): Prevent UnboundLocalError for dataset_alpha when relax_on_silence=False v1.9.47 Hawk Lee 2026-01-22 22:41:50 +08:00
  • c936887aed fix(ditto): Restore EMA Logic! Wire mouth_smoothing to MotionStitch instead of Audio2Motion v1.9.46 Hawk Lee 2026-01-22 22:28:08 +08:00
  • d937f688e3 feat(ditto): Add 'Custom (Manual)' option to resolve conflict with smo_k_d input v1.9.45 Hawk Lee 2026-01-22 22:18:05 +08:00
  • 4125bbce45 feat(ditto): Upgrade mouth_smoothing to Float Blend (0.0/0.3/0.5/0.7) v1.9.44 Hawk Lee 2026-01-22 22:12:56 +08:00
  • aeb2ad027a fix(ditto): Wire up mouth_smoothing parameter (was ignored) v1.9.43 Hawk Lee 2026-01-22 22:08:32 +08:00
  • d4d8103f9f feat(ditto): Implement Auto-Disk Mode (Graph-Aware) v1.9.42; Clean up pose_stabilization Hawk Lee 2026-01-22 21:58:56 +08:00
  • eb2dfb065c fix(ditto): Rollback to v1.9.31 + Add Vitality Noise Injection (v1.9.41) Hawk Lee 2026-01-22 21:32:11 +08:00
  • db6efd07c9 fix(video): Resolve UnboundLocalError and document zero-perf-loss Disk Mode Hawk Lee 2026-01-22 19:35:28 +08:00
  • e7054c2476 fix: Robust filename pattern detection & logging for VideoCombine v1.9.30 Hawk Lee 2026-01-22 19:18:36 +08:00
  • b0a2bf38fa fix: VideoCombine auto-detects filename pattern (fixes %06d vs %08d mismatch) Hawk Lee 2026-01-22 19:11:37 +08:00
  • b3313cdade perf: Maximize I/O with PNG Level 0 (No Compression) Hawk Lee 2026-01-22 19:01:38 +08:00
  • f51c0f5e5b fix: Set default filename_pattern to %08d for seamless Disk Mode integration Hawk Lee 2026-01-22 18:53:01 +08:00
  • 5b0a766908 perf: Optimize Disk Mode speed (Parallel I/O + Fast PNG Compression) Hawk Lee 2026-01-22 18:51:42 +08:00
  • 25557db8e0 fix: Define logger in aiia_ditto_nodes.py to prevent NameError Hawk Lee 2026-01-22 18:39:12 +08:00
  • 5d0ddb56eb docs: Update BodySway documentation with frames_directory input and OOM-Safe workflow Hawk Lee 2026-01-22 18:31:24 +08:00
  • 5e5f7635c0 feat: Add frames_directory input/output to BodySway for OOM-safe disk mode Hawk Lee 2026-01-22 18:30:18 +08:00
  • 391b7967ae docs: Add save_to_disk parameter and frames_dir output documentation Hawk Lee 2026-01-22 18:21:39 +08:00
  • f22abda3b1 feat: Add save_to_disk mode for Ditto to prevent OOM on large frame counts Hawk Lee 2026-01-22 18:19:30 +08:00
  • e6f97e233e perf: Optimize Ditto frame→tensor conversion to avoid memory doubling Hawk Lee 2026-01-22 18:13:00 +08:00
  • 731779be36 perf: True incremental frame release in VideoCombine by converting tensor to list Hawk Lee 2026-01-22 16:21:01 +08:00
  • b9fef53fbb docs: Update README to reflect v1.9.21+ in-memory batch processing optimizations Hawk Lee 2026-01-22 15:58:37 +08:00
  • e0569820ed perf: Optimize AIIA_VideoCombine memory with incremental frame release Hawk Lee 2026-01-22 15:51:27 +08:00
  • 8207326979 perf: Aggressive BodySway memory optimization - smaller batches, explicit del, cuda.empty_cache Hawk Lee 2026-01-22 15:40:37 +08:00
  • b505b521d7 perf: Optimize BodySway memory usage with pre-allocation and batch processing Hawk Lee 2026-01-22 15:10:36 +08:00
  • 2dd439b8e0 fix: Remove EMA reset on speech onset for smooth mouth opening transition Hawk Lee 2026-01-22 14:24:25 +08:00
  • a1ab20588d feat: BodySway v2 - Perlin noise, no Y translation, improved defaults (reduces dizziness) Hawk Lee 2026-01-22 13:51:07 +08:00
  • cbad609b5e fix: Limit BodySway seed to 32-bit for numpy compatibility Hawk Lee 2026-01-22 12:54:44 +08:00
  • 1188bd20e1 refactor: Use global valid region for consistent sway range across all frames Hawk Lee 2026-01-22 12:50:41 +08:00
  • 8c72c6d0c0 feat: Add per-frame runtime boundary check in BodySway to guarantee no black pixels Hawk Lee 2026-01-22 12:48:10 +08:00
  • 3294f2bbd6 fix: Account for rotation margin in BodySway to prevent black corners Hawk Lee 2026-01-22 12:43:44 +08:00
  • 09aa490f00 refactor: Simplify BodySway to use single crop_ratio parameter (auto-calc sway amplitude) Hawk Lee 2026-01-22 12:41:37 +08:00
  • d25026b78e docs: Add section 2.6 Body Sway Post-Processing documentation Hawk Lee 2026-01-22 12:34:27 +08:00
  • 0306b7abdc feat: Add AIIA_BodySway node for simulating subtle body movement via crop/rotation Hawk Lee 2026-01-22 12:29:37 +08:00
  • 9bc12c460c style: Use align attribute for better vertical alignment on GitHub Hawk Lee 2026-01-22 12:16:24 +08:00
  • 7f8d06e1cc style: Combine logo and title on same line using HTML h1 Hawk Lee 2026-01-22 12:05:48 +08:00
  • c5874fa056 feat: Add 48x48 small logo for README header Hawk Lee 2026-01-22 12:04:45 +08:00
  • 8d9fdbd8a0 fix: Use absolute URL for logo in title for better compatibility Hawk Lee 2026-01-22 11:11:49 +08:00
  • bb51d1af9a style: Add scaled logo to README title Hawk Lee 2026-01-22 11:10:46 +08:00
  • 80317d47ae chore: Update icon to use assets/logo.png, fix README formatting Hawk Lee 2026-01-22 09:57:58 +08:00
  • 81872d74e7 chore: Bump version to 1.9.10 (Soft Reset + mouth_smoothing docs) Hawk Lee 2026-01-22 09:52:23 +08:00
  • aecda21338 docs: Add mouth_smoothing docs, remove duplicate mouth_amp entry Hawk Lee 2026-01-22 09:51:19 +08:00
  • 9778d2c654 refactor: Implement Soft Reset - only reset seed, let model continue from idle naturally Hawk Lee 2026-01-22 09:33:40 +08:00
  • d6a4182edd fix: Apply warping at correct frame offset (fuse_r2_s) instead of index 0 Hawk Lee 2026-01-22 01:37:24 +08:00
  • 27800ce3f1 fix: Expand Prediction Warping to ALL kp dimensions (incl. expression) to prevent full reset snap Hawk Lee 2026-01-22 01:16:07 +08:00
  • eb3227a78b fix: Reset EMA on speech onset to prevent stale closed-mouth bleed Hawk Lee 2026-01-22 01:11:43 +08:00
  • eff970828f feat: Expose mouth_smoothing parameter (None/Light/Normal/Heavy) Hawk Lee 2026-01-22 01:06:44 +08:00
  • 50c628f3d6 feat: Add Expression Temporal Smoothing (EMA) to prevent inhumanly fast mouth movements Hawk Lee 2026-01-22 01:02:28 +08:00
  • 0cea1ba9d6 chore: Bump version to 1.9.3 (fix silence_release NameError) Hawk Lee 2026-01-22 00:48:54 +08:00
  • cf3e706dbe fix: NameError - use silence_release arg directly instead of kwargs Hawk Lee 2026-01-22 00:46:40 +08:00
  • 175288f6e9 chore: Bump version to 1.9.2; update description with Ditto performance claims Hawk Lee 2026-01-22 00:45:37 +08:00
  • c1a70dee57 docs: Clarify crop_scale parameter behavior (FOV vs Detail tradeoff) Hawk Lee 2026-01-22 00:37:43 +08:00
  • 6246bac125 docs: Update Ditto Sampler docs with new silence_release and mouth_amp parameters Hawk Lee 2026-01-22 00:36:11 +08:00
  • 7eb6096c4e feat: Add 'silence_release' parameter (Natural/Fast/Deep) to control VAD release speed Hawk Lee 2026-01-22 00:33:34 +08:00
  • b03e838851 style: Tune VAD Release/Silence threshold to 0.8s (Natural Speed) Hawk Lee 2026-01-22 00:26:26 +08:00
  • 8cc1d5fcce feat: Predictive VAD - Only trigger slow release for long silences (>1.4s) to ensure animation completion Hawk Lee 2026-01-22 00:08:11 +08:00
  • 436166b8c2 fix: Bridge short silence gaps (<0.32s) in VAD to prevent premature mouth closing Hawk Lee 2026-01-22 00:02:40 +08:00
  • 832c42c325 fix: Buffer expression history to prevent capturing closed mouth, force full attack strength Hawk Lee 2026-01-21 23:45:19 +08:00
  • 66c021e65f style: Speed up VAD Attack (0.4s -> 0.08s) to fix onset delay, keep Release slow Hawk Lee 2026-01-21 23:31:54 +08:00
  • 1030be5086 fix: Robust handling of speech re-entry (Attack Phase) in MotionStitch Hawk Lee 2026-01-21 22:50:38 +08:00
  • 041fd03036 fix: Prevent instant mouth snap by freezing last speaking frame during VAD release Hawk Lee 2026-01-21 22:44:37 +08:00
  • 0bdd1c165c style: Smooth out VAD envelope (slower attack/release) to prevent mouth snapping Hawk Lee 2026-01-21 22:42:01 +08:00
  • a803a2d9ec revert: Disable Head Pose Relaxation to restore natural drift/vitality Hawk Lee 2026-01-21 22:38:31 +08:00
  • 038a10125a feat: Enable Full Head Pose Relaxation (Expressions + Pose) on Silence Hawk Lee 2026-01-21 22:29:29 +08:00
  • c1cb479775 fix: Replace Smooth Reset with Pre-Fuse Warping (fixes NameError) Hawk Lee 2026-01-21 22:13:00 +08:00
  • 0e5e6e8324 fix: Correct Smooth Reset start index to cover fuse overlap and use Cubic interpolation Hawk Lee 2026-01-21 21:49:45 +08:00
  • e13e9568ea chore: Add debug logging to Audio2Motion to trace Smooth Reset logic Hawk Lee 2026-01-21 21:43:54 +08:00
  • 3a5bf1a1d5 fix: Correct order of setup call in generate to fix UnboundLocalError and ensure blink_amp is passed Hawk Lee 2026-01-21 21:31:48 +08:00
  • 88be1ba587 fix: Remove duplicate argument in MotionStitch.setup causing SyntaxError Hawk Lee 2026-01-21 21:26:58 +08:00
  • b1df5fcd5b fix: Add DITTO_AVAILABLE check to load_model to handle import failures gracefully Hawk Lee 2026-01-21 21:23:42 +08:00
  • a35a069261 feat: Implement Smooth Reset (Pose Blending) to fix head snap at speech onset Hawk Lee 2026-01-21 21:20:00 +08:00
  • b773ee0169 docs: Update README with accurate blink frequency (90-150f) and double-blink status Hawk Lee 2026-01-21 21:17:54 +08:00
  • 04538bebe1 feat: Add blink_amp parameter to control eye closure intensity Hawk Lee 2026-01-21 21:16:14 +08:00
  • 73244a7336 fix: Revert Partial Reset to recover lip sync accuracy (sacrificing smooth head motion) Hawk Lee 2026-01-21 20:51:26 +08:00
  • 12c56c0754 fix: Implement Partial Reset (preserve Pose, reset Exp/Seed) to prevent head snap at speech onset Hawk Lee 2026-01-21 20:40:46 +08:00
  • 7f0ac8e6a8 feat: Tune blink rates to slower human average (10-20bpm / 4-6s interval) Hawk Lee 2026-01-21 18:12:49 +08:00
  • b7da1a19cd fix: Set blink hold duration to fixed 1 frame Hawk Lee 2026-01-21 17:57:05 +08:00
  • cb413354ed feat: Disable Double Blink by default and confirm blink hold duration Hawk Lee 2026-01-21 17:55:37 +08:00
  • eb96aaa966 style: Reduce blink hold duration to 1-3 frames for sharper blink Hawk Lee 2026-01-21 17:47:34 +08:00
  • 0f75bdde56 feat: Implement Blink Hold logic to extend closed-eye duration for more natural realism Hawk Lee 2026-01-21 17:24:30 +08:00
  • ae30694c74 style: Refine blink parameters (lower double blink prob, increase gap) for smoother look Hawk Lee 2026-01-21 17:21:25 +08:00
  • a6faa43b3a feat: Implement Dynamic Blink Rate based on VAD (Fast during speech, Slow during silence) Hawk Lee 2026-01-21 17:16:55 +08:00
  • 5aaf9d8bb0 feat: Tune Normal Blink rate to 50-85 frames (~23bpm) to better match human speaking frequency Hawk Lee 2026-01-21 17:14:06 +08:00
  • 75ef9cb90c docs: Update Ditto Sampler docs with Seed, Double Blink, and Exact Onset Alignment details Hawk Lee 2026-01-21 17:09:53 +08:00
  • a751dd3ac6 feat: Implement Double Blink logic with randomized interval for natural eye motion Hawk Lee 2026-01-21 17:08:31 +08:00
  • 773477d768 fix: Initialize self.seed in ComfyStreamSDK.setup explicitly to avoid AttributeError Hawk Lee 2026-01-21 16:55:30 +08:00
  • 7dfd9197cb fix: Add seed input to AIIA_DittoSampler node definition to resolve NameError Hawk Lee 2026-01-21 16:51:39 +08:00
  • 5320d5048b fix: Implement RNG Seed Reset on Speech Onset to ensure diffusion noise consistency and lip sync accuracy Hawk Lee 2026-01-21 16:44:50 +08:00
  • bf2958b021 fix: Initialize do_reset default value to prevent UnboundLocalError Hawk Lee 2026-01-21 16:19:36 +08:00
  • aeb28f246d fix: Implement Exact Onset Alignment via Variable Step Length to perfectly mimic Clean Start conditions and eliminate jitter Hawk Lee 2026-01-21 16:03:07 +08:00
  • a607adaca0 fix: Reset clip_idx to 0 on onset reset to enforce exact First Segment conditions Hawk Lee 2026-01-21 15:29:26 +08:00
  • 407ae60571 fix: Implement Duration-Based Onset Reset (min 15 frames) to avoid over-resetting on breaths Hawk Lee 2026-01-21 14:49:07 +08:00