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

Commit Graph

  • 0c168e70d6 debug: add diagnostic logging for model loader Hawk Lee 2025-12-31 10:32:57 +08:00
  • 95de2bb900 fix: broaden warning suppression and add runtime model check Hawk Lee 2025-12-31 10:30:15 +08:00
  • 51a781ba2f feat: convert guidance inputs to static text labels using custom JS extension Hawk Lee 2025-12-31 10:28:36 +08:00
  • 52afa11e25 fix: implement robust model switching logic and fix instruction hallucination Hawk Lee 2025-12-31 10:15:49 +08:00
  • b5851715d2 style: add explanatory text labels above TTS input fields for better UI/UX Hawk Lee 2025-12-31 10:04:04 +08:00
  • 8c674f8c43 fix: prioritize RL-Instruct model and fix instruction hallucination issues Hawk Lee 2025-12-31 01:12:55 +08:00
  • 3ae14ba127 chore: suppress noisy PyTorch and ONNX warnings in terminal output Hawk Lee 2025-12-31 01:01:22 +08:00
  • 30af75d8bc feat: enable Pure Instruct Mode for CosyVoice 3.0 via internal neutral seed Hawk Lee 2025-12-31 00:57:10 +08:00
  • 12e589394e fix: resolve KeyError 'pure_1' in CosyVoice 3.0 by adding speaker validation and better defaults Hawk Lee 2025-12-31 00:34:08 +08:00
  • b217dcee89 feat: integrate CosyVoice 3.0 with full capabilities (Instruct, Zero-Shot, Hybrid, SFT, Cross-Lingual) and update documentation Hawk Lee 2025-12-31 00:23:40 +08:00
  • 6ca15fc123 Revert "Standard TTS: Add 'voice' selector and reference audio file support" Hawk Lee 2025-12-30 23:42:08 +08:00
  • 2523dd6ad8 Standard TTS: Add 'voice' selector and reference audio file support Hawk Lee 2025-12-30 23:40:28 +08:00
  • 73bad5abe7 Fix: Remove resilient debug print 'Full Injection' Hawk Lee 2025-12-30 23:30:54 +08:00
  • 6648f3c655 Optimize 0.5B integration: Cleanup debug prints, update docs, mark Preset Maker as experimental Hawk Lee 2025-12-30 23:25:39 +08:00
  • 6614ac07cf Refactor: Separate VibeVoice Realtime TTS and Preset Maker into modular files and enable direct connection Hawk Lee 2025-12-30 16:32:30 +08:00
  • 8cf32887d1 feat: integrate VibeVoice 0.5B realtime support and fix JS deprecation warnings Hawk Lee 2025-12-29 20:43:02 +08:00
  • d4f560cdbb Add support for VibeVoice-Realtime-0.5B (Multi-language & Low latency) Hawk Lee 2025-12-29 19:29:21 +08:00
  • bcedf6b700 Introduce 'auto' sampling mode to fix 1.5B quality while keeping 7B expressiveness Hawk Lee 2025-12-29 19:25:35 +08:00
  • 81f0112094 Smooth progress bar with text-length estimation and final 7B quality fixes Hawk Lee 2025-12-29 19:21:34 +08:00
  • 2ac1d61790 Implement model-specific generation config loading (1.5B vs 7B presets) Hawk Lee 2025-12-29 19:15:20 +08:00
  • ce5182499b Sync defaults with official Microsoft 1.5B config (20 steps, 1.3 CFG) Hawk Lee 2025-12-29 19:12:54 +08:00
  • 521ceb67d7 Optimize 7B model quality with sampling parameters and update README Hawk Lee 2025-12-29 19:10:05 +08:00
  • f49c89d24c Enhance progress bar: jump to 100% on natural termination Hawk Lee 2025-12-29 18:55:29 +08:00
  • 3f2de07ecd Simplify VibeVoice TTS UI: remove max_length_times and set internal default Hawk Lee 2025-12-29 18:53:32 +08:00
  • 9263d3be70 Update README with official VibeVoice parameters and model-specific advice Hawk Lee 2025-12-29 18:48:17 +08:00
  • 7f65bfc749 Finalize VibeVoice integration: 7B support, resampling, punctuation fixes, and docs Hawk Lee 2025-12-29 18:46:12 +08:00
  • 520d7bdcc8 Fix 7B local path detection: use correct org name (vibevoice vs microsoft) Hawk Lee 2025-12-29 18:37:39 +08:00
  • b863cf74e5 Add normalize_text toggle to control text preprocessing Hawk Lee 2025-12-29 18:36:27 +08:00
  • 7f8f700e82 Note 1.5B punctuation limitations, recommend 7B for production Hawk Lee 2025-12-29 18:34:37 +08:00
  • 322ae81333 Make hyphen conversion more conservative: only year ranges with 年 Hawk Lee 2025-12-29 18:34:02 +08:00
  • ac1237977c Add punctuation normalization: replace hyphens with '至', remove quotes Hawk Lee 2025-12-29 18:32:29 +08:00
  • 331cf7ac58 Document CFG scale speed effect and reference audio resampling Hawk Lee 2025-12-29 18:21:43 +08:00
  • d892c60f51 Add automatic resampling of reference audio to 24000Hz Hawk Lee 2025-12-29 18:08:57 +08:00
  • e02e02e0d2 Add debug prints to trace audio length issue Hawk Lee 2025-12-29 18:05:52 +08:00
  • 15d7a317ab Fix 7B download path to vibevoice/VibeVoice-7B Hawk Lee 2025-12-29 17:54:07 +08:00
  • ccfa97d03b Fix 7B path (vibevoice/VibeVoice-7B), add max_length_times param to prevent early cutoff Hawk Lee 2025-12-29 17:52:12 +08:00
  • 121bb6c61a Add speed control parameter (post-process time-stretch) Hawk Lee 2025-12-29 17:47:17 +08:00
  • 9f70ca0155 Add VibeVoice 7B support; update README with language info (en/zh only) Hawk Lee 2025-12-29 17:40:44 +08:00
  • 7f2dd14a1f Fix VibeVoice params: replace language/speed with cfg_scale/ddpm_steps Hawk Lee 2025-12-29 17:35:02 +08:00
  • abf78c0257 Update README: VibeVoice development status and known limitations Hawk Lee 2025-12-29 17:30:47 +08:00
  • c5b9317523 Final cleanup: remove debug prints, delete temp repos Hawk Lee 2025-12-29 17:26:35 +08:00
  • 3e7fee5939 Fix KeyError in non-streaming model: use resolved input_ids Hawk Lee 2025-12-29 17:16:51 +08:00
  • 8172878446 Fix config class mismatch: use non-streaming VibeVoiceConfig for non-streaming model Hawk Lee 2025-12-29 17:12:17 +08:00
  • 8ae16192de Remove temp_repo submodule Hawk Lee 2025-12-29 17:07:48 +08:00
  • be9f4b05f5 Switch to non-streaming VibeVoiceForConditionalGenerationInference Hawk Lee 2025-12-29 17:07:32 +08:00
  • 83072f76fe Pass input_ids positionally to fix transformers generation error Hawk Lee 2025-12-29 16:53:33 +08:00
  • e47d9c432e Refactor VibeVoice generate to handle inputs dynamically and set explicit token IDs Hawk Lee 2025-12-29 16:38:04 +08:00
  • a60810cdca Add debug logging for processor output keys Hawk Lee 2025-12-29 16:17:52 +08:00
  • d2022fc825 Bundle VibeVoice generation_config and implement fallback loading Hawk Lee 2025-12-29 16:10:48 +08:00
  • 848297a804 Fix UnboundLocalError in VibeVoice core Hawk Lee 2025-12-29 15:48:09 +08:00
  • d76d2712e6 Fix VibeVoice load path cleanup bug & update README file structure Hawk Lee 2025-12-29 15:45:16 +08:00
  • e8de3283b3 Refactor VibeVoice to self-contained core (no remote code execution) Hawk Lee 2025-12-29 15:40:35 +08:00
  • e479e26e0a fix: forcefully patch remote model files on disk to bypass AutoModel read-only loading (v1.5.00) Hawk Lee 2025-12-29 15:08:39 +08:00
  • 9468f2d19f fix: use generalized regex for .to(device) runtime patching to match any usage pattern (v1.5.00) Hawk Lee 2025-12-29 15:03:07 +08:00
  • c8e990b9cd fix: correct relative import in bundled vibevoice_processor.py and update task status (v1.5.00) Hawk Lee 2025-12-29 14:59:07 +08:00
  • 5894838e33 fix: skip runtime hot-patching for bundled internal files to avoid double-patching SyntaxErrors (v1.5.00) Hawk Lee 2025-12-29 14:51:09 +08:00
  • 868d04f710 chore: add debug logging to exec() to diagnose SyntaxError from runtime patching (v1.5.00) Hawk Lee 2025-12-29 14:46:35 +08:00
  • b26995f511 fix: use idempotent regex for runtime patching to prevent double-patching existing checks and SyntaxErrors (v1.5.00) Hawk Lee 2025-12-29 14:39:34 +08:00
  • 79d426915b fix: restore runtime safety patching for .to(device) to support unbundled remote model files (v1.5.00) Hawk Lee 2025-12-29 14:36:53 +08:00
  • 45d1c4625a refactor: fully fix all remaining relative imports in vibevoice_core files to absolute imports (v1.5.00) Hawk Lee 2025-12-29 14:28:18 +08:00
  • fbd68ca36c refactor: statically fix relative imports in modeling_vibevoice files to absolute imports to prevent runtime loading errors (v1.5.00) Hawk Lee 2025-12-29 14:15:42 +08:00
  • 7e02593cf7 fix: improved regex for patching relative imports in bundled VibeVoice modules (v1.5.00) Hawk Lee 2025-12-29 14:08:30 +08:00
  • 9a5875d9d7 refactor: replace runtime regex hot-patching with static file fixes in vibevoice_core; prioritize bundled modules (v1.5.00) Hawk Lee 2025-12-29 13:59:02 +08:00
  • b305176b5c fix: pre-patch VibeVoice dependencies (modeling_vibevoice_inference) to ensure safety patches are applied recursively (v1.5.00) Hawk Lee 2025-12-29 13:45:34 +08:00
  • a532d6a63d fix: add missing 'schedule' dependency to vibevoice_core for modeling_vibevoice.py (v1.5.00) Hawk Lee 2025-12-29 13:36:48 +08:00
  • e5384ae67c fix: reliably load and register VibeVoice sub-model classes by falling back to modeling_vibevoice.py (v1.5.00) Hawk Lee 2025-12-29 13:29:42 +08:00
  • 68413406f7 fix: register all VibeVoice sub-model classes (Acoustic, Semantic, DiffusionHead) with AutoModel (v1.5.00) Hawk Lee 2025-12-29 13:18:53 +08:00
  • 2b54f1a99b fix: remove incorrect registration of VibeVoiceStreamingConfig to resolve model_type mismatch (v1.5.00) Hawk Lee 2025-12-29 12:33:18 +08:00
  • 3e828a59d3 fix: register all nested VibeVoice config classes (Acoustic, Semantic, etc.) to AutoConfig (v1.5.00) Hawk Lee 2025-12-29 11:29:27 +08:00
  • 5d4107313e fix: force hot-patching of VibeVoice modules even if cached, with robust regex (v1.5.00) Hawk Lee 2025-12-29 11:12:42 +08:00
  • aaa4ec5adf fix: improved regex for hot-patching VibeVoice model code to be more robust (v1.5.00) Hawk Lee 2025-12-29 11:04:41 +08:00
  • 5759fb9087 fix: use robust regex hot-patching for external VibeVoice code (v1.5.00) Hawk Lee 2025-12-29 10:33:35 +08:00
  • 81550c923e fix: hot-patch loaded model code to prevent crashes on None inputs (v1.5.00) Hawk Lee 2025-12-29 10:11:57 +08:00
  • fbeea41edd fix: enforce reference_audio requirement to prevent model inference crash (v1.5.00) Hawk Lee 2025-12-29 10:07:20 +08:00
  • e067dd7cf1 fix: safely handle None optional inputs in VibeVoice generation (v1.5.00) Hawk Lee 2025-12-29 10:03:47 +08:00
  • 799e639a55 fix: monkey-patch tokenizer with VibeVoice specific attributes (v1.5.00) Hawk Lee 2025-12-29 10:00:26 +08:00
  • 6f7f1dd993 fix: auto-format text input with speaker tags if missing (v1.5.00) Hawk Lee 2025-12-29 09:57:52 +08:00
  • 2e5ae0bf9e fix: prevent repeated module loading to avoid type mismatch in AutoConfig registry (v1.5.00) Hawk Lee 2025-12-29 09:55:59 +08:00
  • 5c7148b147 fix: add missing vibevoice_core files and update sys.path for relative imports (v1.5.00) Hawk Lee 2025-12-29 09:49:32 +08:00
  • 49337afe25 fix: remove nested try block in v1.5.00 Hawk Lee 2025-12-29 09:39:57 +08:00
  • e0bddc23ae fix: resolve syntax error and clean up module loading in v1.5.00 Hawk Lee 2025-12-29 09:33:05 +08:00
  • 7efe3d66fe fix: resolve code duplication and try-except syntax in v1.5.00 Hawk Lee 2025-12-29 09:29:12 +08:00
  • 5140148dd8 chore: bump version to 1.4.99 Hawk Lee 2025-12-29 01:18:35 +08:00
  • 6fb89b083e feat: implement VibeVoiceProcessor integration and dynamic loading (v1.4.99) Hawk Lee 2025-12-29 01:17:43 +08:00
  • f65cda6c2f fix: pass tokenizer to VibeVoice generate (v1.4.98) Hawk Lee 2025-12-29 01:13:37 +08:00
  • 89a508cec6 docs: update VibeVoice instructions with Qwen tokenizer details Hawk Lee 2025-12-29 01:09:58 +08:00
  • 0e50c02848 fix: implement Qwen2Tokenizer fallback for VibeVoice (v1.4.97) Hawk Lee 2025-12-29 01:06:01 +08:00
  • 5ad35854e6 fix: clean overwrite of aiia_vibevoice_nodes.py (v1.4.96) Hawk Lee 2025-12-29 01:00:29 +08:00
  • 4b0b438f9b fix: refactor load_model to correct indentation (v1.4.96) Hawk Lee 2025-12-29 00:55:31 +08:00
  • 3065aea44b fix: bypass AutoTokenizer with explicit class loading (v1.4.95) Hawk Lee 2025-12-29 00:52:43 +08:00
  • c76dcd543e fix: retry register VibeVoice tokenizer mapping (v1.4.94) Hawk Lee 2025-12-29 00:47:56 +08:00
  • 5068594590 fix: register VibeVoice tokenizer mapping (v1.4.94) Hawk Lee 2025-12-29 00:47:30 +08:00
  • ffd1786f8b fix: reconcile model_type to 'vibevoice' (v1.4.93) Hawk Lee 2025-12-29 00:41:58 +08:00
  • be842000e3 fix: reconcile model_type mismatch for VibeVoice 1.5B (v1.4.93) Hawk Lee 2025-12-29 00:41:18 +08:00
  • 87c7c4e81c fix: add missing imports re/types for VibeVoice patching (v1.4.92) Hawk Lee 2025-12-29 00:24:29 +08:00
  • 7f7287542a fix: dynamic source patching to resolve relative imports in VibeVoice (v1.4.91) Hawk Lee 2025-12-29 00:19:32 +08:00
  • d00477862a fix: alias 0.5B code to support 1.5B VibeVoice loading (v1.4.90) Hawk Lee 2025-12-29 00:13:06 +08:00
  • 8db4823daf fix: use importlib for robust VibeVoice loading (v1.4.89) Hawk Lee 2025-12-28 23:56:40 +08:00
  • 3fa162f657 fix: manual registration of VibeVoice arch to bypass missing auto_map (v1.4.88) Hawk Lee 2025-12-28 23:55:13 +08:00
  • ebf0c25f6f fix: add deep debugging for VibeVoice loader (v1.4.87) Hawk Lee 2025-12-28 23:49:05 +08:00