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

Commit Graph

  • 31c7b08a6b fix(wan): ensure input tensor is on correct device in patch_embed Hawk Lee 2026-01-20 00:08:37 +08:00
  • 53ac7dc881 fix(wan): disable autocast for patch_embedding Hawk Lee 2026-01-20 00:06:04 +08:00
  • 5254567c32 fix(wan): force F.conv3d with explicit float32 weights Hawk Lee 2026-01-20 00:03:18 +08:00
  • f9f1c58cc7 fix(test): correct encode_prompt arguments Hawk Lee 2026-01-20 00:00:49 +08:00
  • 28d22fb086 chore: add final verification script Hawk Lee 2026-01-19 23:57:55 +08:00
  • 798d742a0f fix(test): add missing torchaudio import Hawk Lee 2026-01-19 23:55:41 +08:00
  • b6d9a925a2 chore: add version tag to test script Hawk Lee 2026-01-19 23:54:07 +08:00
  • 8101c25d81 fix(test): use explicit pth paths for encoders Hawk Lee 2026-01-19 23:52:40 +08:00
  • caa24eec2a chore: add test_server_fix.py with correct paths Hawk Lee 2026-01-19 23:51:21 +08:00
  • cfdfd564ba fix(test): correct model paths in server verification script Hawk Lee 2026-01-19 23:49:31 +08:00
  • 7d42466b38 chore: add manual test script for server verification Hawk Lee 2026-01-19 23:45:52 +08:00
  • fa50a5dae2 fix(transformer): use in-place float32 conversion for patch_embedding Hawk Lee 2026-01-19 23:43:44 +08:00
  • f1848bb81a fix(transformer): disable autocast for forced float32 conv3d Hawk Lee 2026-01-19 23:38:14 +08:00
  • 67cc60515b fix(transformer): force patch_embedding to float32 Hawk Lee 2026-01-19 23:33:20 +08:00
  • 09292cb631 fix(transformer): add .contiguous() to patch_embedding input Hawk Lee 2026-01-19 23:26:04 +08:00
  • ccd416ad2a fix(pipeline): ensure correct device usage in loop Hawk Lee 2026-01-19 23:22:18 +08:00
  • 7c14df3fbe fix(pipeline): fix indentation bug in prepare_mask_latents Hawk Lee 2026-01-19 23:13:29 +08:00
  • 220e1146b9 fix(echomimic): add .contiguous() to vae and move back to GPU Hawk Lee 2026-01-19 23:10:12 +08:00
  • 312aa234f0 feat(debug): add logging to VAE encoding loop Hawk Lee 2026-01-19 23:08:33 +08:00
  • 130ae46f86 fix(echomimic): run VAE on CPU to avoid cuda 3D conv error Hawk Lee 2026-01-19 22:51:47 +08:00
  • b1df960d51 fix(echomimic): force VAE to float32 to fix slow_conv3d_forward error on cuda Hawk Lee 2026-01-19 22:19:19 +08:00
  • 59ff8bd825 fix(echomimic): restore missing ip_mask definition Hawk Lee 2026-01-19 22:05:14 +08:00
  • 3a15acebc8 fix(echomimic): use cpu generator to avoid device mismatch with offloaded text encoder Hawk Lee 2026-01-19 22:01:00 +08:00
  • 5728e75829 fix(pipeline): add clip_context argument to __call__ Hawk Lee 2026-01-19 21:48:02 +08:00
  • 97328f2d90 fix(echomimic): add missing TF import Hawk Lee 2026-01-19 21:45:08 +08:00
  • 10714a2e56 fix(segments): add aggressive vram cleanup for nemo model Hawk Lee 2026-01-19 21:41:30 +08:00
  • 7bc72eca9e fix(audio): aggressive vram cleanup for vibevoice and audio_enhance nodes Hawk Lee 2026-01-19 21:39:56 +08:00
  • d75303b1b6 perf(echomimic): pre-compute CLIP context to save VRAM Hawk Lee 2026-01-19 21:35:03 +08:00
  • 317b693a4a perf(echomimic): permit passing pre-computed clip_context to pipeline Hawk Lee 2026-01-19 21:34:26 +08:00
  • c814546fb0 perf(echomimic): add post-generation model cleanup (move to CPU) Hawk Lee 2026-01-19 21:26:01 +08:00
  • cd87d26437 perf(echomimic): implement manual device management for robust speed optimization Hawk Lee 2026-01-19 21:25:28 +08:00
  • 741a95909a fix(echomimic): handle missing text_encoder in pipeline to support memory optimization Hawk Lee 2026-01-19 21:17:54 +08:00
  • aa67ce4b13 perf(echomimic): delete text_encoder after use to free 10GB VRAM & remove debug prints Hawk Lee 2026-01-19 21:11:20 +08:00
  • cf0775785d debug: add print statements to pipeline Hawk Lee 2026-01-19 21:05:57 +08:00
  • cd701b0032 fix(echomimic): correct batch_size logic in pipeline __call__ method Hawk Lee 2026-01-19 21:01:11 +08:00
  • d51c6c07e9 fix(echomimic): correct batch_size logic (take 2) Hawk Lee 2026-01-19 20:50:46 +08:00
  • 5f48fd666c fix(echomimic): revert transpose and implement correct flat-list CFG Hawk Lee 2026-01-19 20:41:19 +08:00
  • 13708df19c fix(echomimic): transpose T5 embeddings to Time-Major [S,B,D] for transformer Hawk Lee 2026-01-19 20:31:36 +08:00
  • d624a8e06f fix(echomimic): correct I2V context concatenation for 4D T5 layers Hawk Lee 2026-01-19 20:26:04 +08:00
  • dd0a41ae72 fix(echomimic): correct CFG embedding concatenation for list inputs Hawk Lee 2026-01-19 20:20:26 +08:00
  • e9e5141a0c fix(echomimic): syntax error in pipeline Hawk Lee 2026-01-19 20:10:13 +08:00
  • adb2e780d9 fix(echomimic): handle list-type prompt_embeds in encode_prompt Hawk Lee 2026-01-19 20:09:55 +08:00
  • d928adb501 fix(echomimic): add mm import Hawk Lee 2026-01-19 20:03:15 +08:00
  • 5aaf22b20a fix(echomimic): handle list embeddings in pipeline and verbose memory cleanup Hawk Lee 2026-01-19 20:00:52 +08:00
  • 4733cc1685 fix(echomimic): add missing pre-encoding logic Hawk Lee 2026-01-19 19:57:27 +08:00
  • 2d7926c2d4 perf(echomimic): pre-encode prompts and unload models to fix slowness Hawk Lee 2026-01-19 19:52:25 +08:00
  • 951c52c30c fix(echomimic): reorder debug print and add process memory log Hawk Lee 2026-01-19 19:43:30 +08:00
  • 2956f683ca fix(echomimic): resolve device mismatch in blending and debug memory Hawk Lee 2026-01-19 19:39:44 +08:00
  • 2b60e65d07 docs(echomimic): restore Wan2.1 download instructions Hawk Lee 2026-01-19 18:50:24 +08:00
  • b9bd2cf65e fix(echomimic): fallback to base model for missing components Hawk Lee 2026-01-19 18:49:25 +08:00
  • f9cdac4414 fix(echomimic): update config subpaths to standard folder names Hawk Lee 2026-01-19 18:46:48 +08:00
  • b4ca6d0d35 docs(echomimic): update model source to BadToBest/EchoMimicV3 Hawk Lee 2026-01-19 18:39:39 +08:00
  • c580668329 fix(echomimic): disable TF gpu and use sequential offload Hawk Lee 2026-01-19 18:18:37 +08:00
  • 0359000f5a fix(echomimic): force models to CPU during loading to fix OOM Hawk Lee 2026-01-19 18:08:59 +08:00
  • 6f7010394a chore(echomimic): add debug logging for vram usage Hawk Lee 2026-01-19 18:04:48 +08:00
  • 36aa61cbc5 fix(echomimic): reduce default context_length to 49 and expose it as param Hawk Lee 2026-01-19 18:00:05 +08:00
  • e5808483e4 fix(echomimic): implement sliding window chunking to fix OOM Hawk Lee 2026-01-19 17:55:51 +08:00
  • 6b95605acc fix(echomimic): resolve dtype mismatch in wav2vec between input and weights Hawk Lee 2026-01-19 17:52:12 +08:00
  • d2e8d8da8b fix(echomimic): enable model cpu offload to prevent OOM Hawk Lee 2026-01-19 17:48:21 +08:00
  • c0ef400007 fix(echomimic): update model loading logic to use config.yaml Hawk Lee 2026-01-19 17:45:17 +08:00
  • f45d3c0984 fix(echomimic): force add dist folder in new location Hawk Lee 2026-01-19 17:42:13 +08:00
  • dc1acb8c0e refactor(echomimic): rename src to echomimic_v3_src to avoid namespace collisions Hawk Lee 2026-01-19 17:40:27 +08:00
  • d88eacc777 fix(echomimic): force add missing __init__.py Hawk Lee 2026-01-19 17:38:16 +08:00
  • 48f28396bd fix(echomimic): restore broken code and improve import error reporting Hawk Lee 2026-01-19 17:31:39 +08:00
  • 351111eeeb fix: force add src/dist module Hawk Lee 2026-01-19 17:24:58 +08:00
  • 7c84babf6d docs: use 'hf download' and fix wav2vec2 instruction Hawk Lee 2026-01-19 16:53:55 +08:00
  • a0454dd677 docs: standardize download commands to huggingface-cli Hawk Lee 2026-01-19 16:52:00 +08:00
  • 76ce044e02 docs: use huggingface-cli for model download instructions Hawk Lee 2026-01-19 16:50:47 +08:00
  • 982a1625a2 docs: update Wan2.1 model link in README Hawk Lee 2026-01-19 16:48:52 +08:00
  • e8df949eb8 docs: standardize numbering for video generation nodes Hawk Lee 2026-01-19 16:45:15 +08:00
  • 099eb9033c docs: add EchoMimic V3 setup instructions Hawk Lee 2026-01-19 16:43:33 +08:00
  • c9441f1b92 feat: add EchoMimic V3 nodes Hawk Lee 2026-01-19 16:40:13 +08:00
  • b095878500 fix(float): change mask_top_edge_pixels to crop pixels instead of smear Hawk Lee 2026-01-19 15:54:40 +08:00
  • 2661577b5a Revert "fix: revert video artifact fix causing deformation" Hawk Lee 2026-01-19 15:46:40 +08:00
  • 5eacb02093 fix: revert video artifact fix causing deformation Hawk Lee 2026-01-19 15:41:29 +08:00
  • 8068503037 feat: add mask_top_edge_pixels to float nodes to fix video artifacts Hawk Lee 2026-01-19 15:33:06 +08:00
  • 3af5060421 chore: bump version to 1.8.4 for registry release Hawk Lee 2026-01-19 13:44:49 +08:00
  • ac701b56ee docs: update changelog and prerequisites for v1.8.4 (sox support) Hawk Lee 2026-01-19 13:43:51 +08:00
  • 181b6d9b20 fix: use '-s' flag in sox tempo for better quality Hawk Lee 2026-01-19 13:39:41 +08:00
  • 83b4f22be7 fix: normalize audio before sox tempo to prevent clipping Hawk Lee 2026-01-19 13:30:23 +08:00
  • ea606e8224 fix: unconditionally cast audio output to float32 Hawk Lee 2026-01-19 13:20:06 +08:00
  • db58411b8a feat: add log for successful time stretch Hawk Lee 2026-01-19 13:15:57 +08:00
  • e48de74ee3 fix: use system sox command via subprocess for robust time stretching Hawk Lee 2026-01-19 13:06:02 +08:00
  • ae1864a57b fix: VibeVoice speed control and add sox dependency Hawk Lee 2026-01-19 12:51:19 +08:00
  • 1ac8d65f01 docs: Update README for VibeVoice optional reference audio (v1.8.3) Hawk Lee 2026-01-07 10:49:23 +08:00
  • 279527e29c feat: Make reference_audio optional for VibeVoice TTS (Standard) with auto-fallback to default seed (v1.8.3) Hawk Lee 2026-01-07 10:47:24 +08:00
  • 5ee0a975d2 chore: Bump version to 1.8.2 to trigger registry fix verification Hawk Lee 2026-01-06 20:04:37 +08:00
  • 7d8fe8ea4a fix: Remove nested submodule references from libs/CosyVoice Hawk Lee 2026-01-06 20:01:17 +08:00
  • 22bce8dd7d fix: Improve CosyVoice error messages to show actual failure reason Hawk Lee 2026-01-06 19:56:37 +08:00
  • 9894410fff docs: Update README with Interactive Teaching (Web Export) and Caching Optimization (v1.8.1) Hawk Lee 2026-01-05 15:31:47 +08:00
  • 6f5446c20e feat: Workflow Caching Optimization (v1.8.1) - Modify Script Parser to return clean tts_data (enabling TTS caching) - Add AIIA_Segment_Merge node to re-attach visual tags post-TTS - Update docs with Separate Streams workflow Hawk Lee 2026-01-05 15:15:35 +08:00
  • 8d933f0205 feat: Interactive Teaching Web Export (v1.8.0) - Add AIIA_Web_Export node for generating HTML player - Update Script Parser to support (Visual: url) tags - Update Dialogue TTS to propagate visual tags to segments - Support relative paths for visual content Hawk Lee 2026-01-05 15:07:38 +08:00
  • 8f566dc16e fix: Generate per-speaker ASS styles with distinct colors (v1.7.2) - Automatically assign colors to speakers in ASS output - Fixes issue where all speakers used default white style Hawk Lee 2026-01-05 14:01:59 +08:00
  • 887d99c5db docs: Update README with AIIA_Subtitle_Preview node Hawk Lee 2026-01-05 13:48:16 +08:00
  • 4c34904cf4 feat: Add AIIA_Subtitle_Preview node (v1.7.1) - New node for previewing audio with synchronized subtitles - Supports SRT and basic ASS styles (Color, Font Size) - Includes Javascript frontend for timeline playback Hawk Lee 2026-01-05 13:46:15 +08:00
  • 57285ec603 feat: Add Subtitle Generation support (v1.7.0) - Add AIIA_Subtitle_Gen node for SRT/ASS generation - Update AIIA_Dialogue_TTS to output precise timestamps (segments_info) - Implement Smart Batch Interpolation for VibeVoice timestamps - Update README and Walkthrough Hawk Lee 2026-01-05 13:33:57 +08:00
  • 4cccc249dc chore: cleanup unused assets replaced by user attachment Hawk Lee 2026-01-05 12:38:08 +08:00
  • 97365676e2 Update README.md Hawk 2026-01-05 12:36:23 +08:00
  • 3927dc9a66 Update README.md Hawk 2026-01-05 12:34:56 +08:00
  • eca7d8438f docs: use raw video link as fallback Hawk Lee 2026-01-05 12:29:42 +08:00