This website requires JavaScript.
77eb7f63d0
docs: use image link to absolute blob url for safe video playback
Hawk Lee
2026-01-05 12:27:28 +08:00
d124805073
docs: try html5 video tag for inline playback
Hawk Lee
2026-01-05 12:25:50 +08:00
1bda0870d3
docs: use video demo for reliable playback support
Hawk Lee
2026-01-05 12:23:31 +08:00
53d6d57f34
docs: use absolute raw url for audio playback support
Hawk Lee
2026-01-05 12:21:18 +08:00
e9474b85a1
docs: replace html audio tag with markdown link for github compatibility
Hawk Lee
2026-01-05 12:19:23 +08:00
1badf2f8a4
docs: embed podcast audio sample and screenshots in README
Hawk Lee
2026-01-05 12:17:39 +08:00
e76af6119e
fix: resolve submodule error by vendoring CosyVoice lib + bump version to 1.6.2
Hawk Lee
2026-01-04 23:41:10 +08:00
673432e628
chore: bump version to 1.6.1 and enrich metadata for Comfy registry
Hawk Lee
2026-01-04 23:38:08 +08:00
ae9c663fc4
chore: bump version to 1.6.0 for Podcast/Dialogue system
Hawk Lee
2026-01-04 20:29:21 +08:00
174cc81609
feat(podcast): Add VibeVoice Batch Modes and Fixes - Add Batch Mode (Natural, Strict, Whole) to Dialogue TTS - Fix VibeVoice trailing noise with auto-trim - Fix built-in audio clipping by attenuating volume - Add comment support to Script Parser - Update README with new features
Hawk Lee
2026-01-04 20:27:06 +08:00
5ea20cb94f
Update README
Hawk Lee
2026-01-04 18:17:57 +08:00
68d5df99d8
Fix: Remove explicit resampling logic from VoxCPM node (model is now native 44.1k)
Hawk Lee
2026-01-04 18:00:14 +08:00
f9bff68f07
Fix: Hardcode AudioVAE default params to 44.1kHz (VoxCPM 1.5 Native) to bypass config loading issues
Hawk Lee
2026-01-04 17:42:31 +08:00
ba9a62267c
Refactor: Fully replace vendored VoxCPM core with Wildminder's High-Fidelity implementation (src/voxcpm)
Hawk Lee
2026-01-04 17:32:17 +08:00
2e0a00f56b
Feat: Resample output to 44.1kHz to meet user expectations (upsampling 16k->44.1k)
Hawk Lee
2026-01-04 17:09:32 +08:00
de6b1084dc
Revert: Restore 44.1kHz sample rate as user confirmed file spectrum is correct (fast speed is pace, not pitch)
Hawk Lee
2026-01-04 17:04:59 +08:00
ef3a9b6320
Fix: Force 16kHz output rate to resolve fast playback speed (despite 44.1k config claim)
Hawk Lee
2026-01-04 16:58:33 +08:00
e2b235ff43
Fix: correct sample rate from model metadata (fixes fast playback) and remove debug prints
Hawk Lee
2026-01-04 16:44:53 +08:00
edeb239efa
Debug: Add print statements to trace audio tensor shape issues
Hawk Lee
2026-01-04 16:39:41 +08:00
a25ba3946c
Fix: Output 3D audio tensor [Batch, Channel, Time] for ComfyUI compatibility
Hawk Lee
2026-01-04 16:37:04 +08:00
95c60998b6
Fix: Correct indentation of try-except block in generate method
Hawk Lee
2026-01-04 16:33:38 +08:00
2bb1633e0f
Fix: Indentation error in generate method
Hawk Lee
2026-01-04 16:30:33 +08:00
a79d308030
Fix: Restore missing try-block causing SyntaxError
Hawk Lee
2026-01-04 16:28:56 +08:00
0f214d10bf
Fix: Ensure prompt_text is None if no audio prompt is provided (fixes VoxCPM strict validation)
Hawk Lee
2026-01-04 16:27:00 +08:00
6d2483a983
Fix: Align VoxCPM generate arguments (rename prompt_wav_path, add prompt_text, remove unsupported params)
Hawk Lee
2026-01-04 16:23:05 +08:00
e3a6b4c721
Fix: Disable torch.compile (optimize=False) to resolve cudaMallocAsync crash
Hawk Lee
2026-01-04 16:20:53 +08:00
6ae43bc55d
Fix(UI): Remove long labels from enable_denoiser to prevent text overlap
Hawk Lee
2026-01-04 16:19:09 +08:00
7b258fee31
Docs: Update ZipEnhancer path to local models dir
Hawk Lee
2026-01-04 16:15:50 +08:00
2fdb460f26
Fix: Download ZipEnhancer to local models dir instead of ~/.cache
Hawk Lee
2026-01-04 16:15:43 +08:00
6b50b1744c
Docs: Add ZipEnhancer manual download instructions
Hawk Lee
2026-01-04 16:14:18 +08:00
de20148036
Feat: Add 'enable_denoiser' toggle to VoxCPM Loader
Hawk Lee
2026-01-04 16:13:10 +08:00
e04e0b907e
Fix: Restore base_path definition for locating core module
Hawk Lee
2026-01-04 16:01:32 +08:00
ad766435d2
Fix: Remove 'device' arg from VoxCPM init (not supported by core)
Hawk Lee
2026-01-04 15:58:37 +08:00
5d122b74c0
Refactor: Vendor VoxCPM core code to avoid pip dependency hell
Hawk Lee
2026-01-04 15:55:59 +08:00
b4ac7603dd
Fix: Remove auto-install to prevent dependency hell (VoxCPM requires Torch>=2.5)
Hawk Lee
2026-01-04 15:51:52 +08:00
bd3ba6caa8
Feat: Auto-install voxcpm dependency if missing
Hawk Lee
2026-01-04 15:50:09 +08:00
ca6181a6df
Fix: Switch to 'voxcpm' python package for loading
Hawk Lee
2026-01-04 15:49:03 +08:00
2b66f12ac2
Fix: Move VoxCPM model path to global ComfyUI/models dir
Hawk Lee
2026-01-04 15:46:22 +08:00
6ab48f0adc
Feat: Implement VoxCPM Inference Logic
Hawk Lee
2026-01-04 15:43:24 +08:00
392fb5dd9c
Docs: Fix VoxCPM Repo ID in README
Hawk Lee
2026-01-04 15:38:41 +08:00
5c979556a9
Fix: Update VoxCPM Repo ID to openbmb/VoxCPM1.5
Hawk Lee
2026-01-04 15:38:34 +08:00
8f1cde850e
Docs: Add VoxCPM 1.5 manual download instructions
Hawk Lee
2026-01-04 15:35:36 +08:00
d0d652446b
Feat: Add VoxCPM 1.5 Node Skeleton
Hawk Lee
2026-01-04 15:34:42 +08:00
6cfa1274e2
Docs: Add Voice Conversion support matrix (CosyVoice wins Audio-2-Audio)
Hawk Lee
2026-01-04 15:27:22 +08:00
00ee3c0b35
Docs: Add VoxCPM/CosyVoice/VibeVoice detailed comparison table
Hawk Lee
2026-01-04 15:20:30 +08:00
3147e5e997
Docs: Add context length comparison (Unlimited Streaming) to Sec 3.5
Hawk Lee
2026-01-04 15:12:02 +08:00
185bab0507
Release: v1.5.0 - Unified CosyVoice Fix & Updated Docs
Hawk Lee
2026-01-04 15:03:53 +08:00
9c0228af5c
Docs: Correct language recommendation to English only
Hawk Lee
2026-01-04 15:00:07 +08:00
c26ca18f20
Docs: Clarify language support for 300M-Instruct
Hawk Lee
2026-01-04 14:56:27 +08:00
55afa827ea
Fix: 300M-Instruct instruction following via <|endofprompt|> and update README
Hawk Lee
2026-01-04 14:50:26 +08:00
0913266ea2
Fix: Protect <|endofprompt|> from normalization in V1 Instruct
Hawk Lee
2026-01-04 12:54:08 +08:00
bad3b04619
Fix: Final restoration of 300M-Instruct (Native Surgical Path + Correct Boundary Tokens)
Hawk Lee
2026-01-04 00:30:31 +08:00
fff0f1bd11
Fix: Restored V1 instruction assembly syntax and switched to native inference_instruct
Hawk Lee
2026-01-04 00:23:26 +08:00
b3a6f4ccf2
Fix: Corrected spk2info key and refined V1 Instruct identity injection
Hawk Lee
2026-01-03 23:58:31 +08:00
56f51b4217
Fix: Final restoration for 300M-Instruct (boundary tokens and identity preservation)
Hawk Lee
2026-01-03 23:51:22 +08:00
22e2d9784a
Fix: Ensure 300M-Instruct uses the surgical instruction path in identity-only mode
Hawk Lee
2026-01-03 23:40:08 +08:00
c9df64d283
Assets: Added high-quality female and male seeds for CosyVoice V2/V3
Hawk Lee
2026-01-03 23:25:32 +08:00
fb68fa6421
Optimize: V3 seed quality and V1 Instruct following
Hawk Lee
2026-01-03 20:01:50 +08:00
584d8e234d
Baseline: Unified CosyVoice V1/V2/V3 support and stability fixes
Hawk Lee
2026-01-03 19:36:07 +08:00
b6b3b09b57
fix: restore 100% native paths for 300M models and remove hacks
Hawk Lee
2026-01-01 16:19:58 +08:00
c062cac0ba
fix: restore native 300M performance and tone quality
Hawk Lee
2026-01-01 16:14:04 +08:00
f65b561e9b
fix: resolve structural indentation mismatch in generate function
Hawk Lee
2026-01-01 15:57:12 +08:00
c8493beb4d
fix: resolve indentation syntax error in inference block
Hawk Lee
2026-01-01 15:46:03 +08:00
a6cc297824
fix: implement native Zero-Shot Fusion strategy for 300M series
Hawk Lee
2026-01-01 15:32:19 +08:00
c8698a8652
fix: relax spk_id validation for 300M Base model
Hawk Lee
2026-01-01 15:06:06 +08:00
784555192f
fix: move types import to global scope
Hawk Lee
2026-01-01 15:01:20 +08:00
c0d0872401
fix: refine V1 speed boost and strengthen identity injection
Hawk Lee
2026-01-01 01:29:22 +08:00
bdc3aad486
fix: implement Identity Injection and speed boost for 300M series
Hawk Lee
2026-01-01 01:26:07 +08:00
107405f8db
fix: revert to native routing for 300M series, fix base model babbling
Hawk Lee
2026-01-01 01:03:51 +08:00
66bee30afa
fix: universal zero-shot routing for 300M-Instruct male voice
Hawk Lee
2026-01-01 00:52:30 +08:00
a6fc576d2b
fix: restore identity monkeypatch and sync deep male seed
Hawk Lee
2026-01-01 00:42:07 +08:00
7851399758
fix: definitive 300M series fix with descriptive hints and PCM_16
Hawk Lee
2026-01-01 00:27:31 +08:00
c4b8efe53d
fix: final 300M series routing and hq seed fix
Hawk Lee
2025-12-31 23:46:03 +08:00
bd0a2ccdb6
fix: definitive 300M routing and hq male seed
Hawk Lee
2025-12-31 21:40:14 +08:00
1d2b4c6158
fix: 300M definitive fix for gender, volume and routing
Hawk Lee
2025-12-31 20:55:18 +08:00
2186323fba
fix: 300M series definitive fixes (gender, volume, dialects, and cutoff)
Hawk Lee
2025-12-31 20:44:04 +08:00
f6a733ce69
fix: resolve 300M series quirks (final routing and audio processing)
Hawk Lee
2025-12-31 19:13:03 +08:00
876269de2a
fix: resolve 300M gender, volume, and muffled spectrogram issues
Hawk Lee
2025-12-31 18:49:42 +08:00
e8fe06e872
fix: resolve 300M quirks (reading prompts aloud and low volume)
Hawk Lee
2025-12-31 18:35:43 +08:00
bd3d4f134c
fix: refactor generate method to centralize instructions and resolve V1 Instruct gender issues
Hawk Lee
2025-12-31 13:58:19 +08:00
0cd0d394c6
fix: resolve 300M-Instruct gender issues via text hints and 300M-Base content via correct male seed transcript
Hawk Lee
2025-12-31 13:52:27 +08:00
9f8f875539
fix: resolve V1 LLM sampling error by avoiding mismatched prompt text; add .txt transcript support
Hawk Lee
2025-12-31 13:47:11 +08:00
d18b62589d
fix: refine model version detection to prioritize yaml config over flow.pt presence
Hawk Lee
2025-12-31 13:33:37 +08:00
0d1391609f
fix: resolve STFT window mismatch error via Matcha cache clearing and robust seed resampling; improve V2 detection
Hawk Lee
2025-12-31 13:02:07 +08:00
81a646c41f
fix: restore base_gender functionality for 300M models (SFT auto-select and Base prompt text)
Hawk Lee
2025-12-31 12:35:18 +08:00
3987b162ef
refactor: rename base_timbre to base_gender for better UX
Hawk Lee
2025-12-31 12:04:26 +08:00
5a8370f272
docs: clarify 0.5B Instruct mode as style/emotion modulation
Hawk Lee
2025-12-31 11:54:44 +08:00
d026d53891
fix: resolve KeyError: '' for V1 models by improving inference path selection and spk_id fallbacks
Hawk Lee
2025-12-31 11:48:40 +08:00
cd2bf91edc
docs: add 300M series model details and limitations to README
Hawk Lee
2025-12-31 11:46:05 +08:00
92223a9a07
fix: add safety check for ttsfrd and missing model configs
Hawk Lee
2025-12-31 11:41:33 +08:00
47df796deb
fix: remove CosyVoice-ttsfrd from model list as it is a dependency, not a model
Hawk Lee
2025-12-31 11:41:10 +08:00
9c6ce0ca49
docs: add comprehensive model download CLI commands to README
Hawk Lee
2025-12-31 11:38:30 +08:00
713ed5feae
fix: merge dialect and emotion into a single instruction sentence to prevent conflict
Hawk Lee
2025-12-31 11:30:45 +08:00
73b71fad34
feat: refine preset labels and add tooltips for conflict handling
Hawk Lee
2025-12-31 11:26:49 +08:00
9d7f4a6071
chore: disable debug and initialization logs
Hawk Lee
2025-12-31 11:25:46 +08:00
e35e28b6aa
docs: update README with CosyVoice 3.0 new parameters and modes
Hawk Lee
2025-12-31 11:18:56 +08:00
0ae678f352
feat: implement dialect and emotion presets for CosyVoice TTS
Hawk Lee
2025-12-31 11:17:07 +08:00
d13af58e46
feat: add base_timbre selection to support gender switching in Instruct mode
Hawk Lee
2025-12-31 11:06:28 +08:00
0ada06fe63
fix: automate V3/V2 instruction formatting with system prompt and endofprompt separator
Hawk Lee
2025-12-31 11:01:02 +08:00
b6b24186dc
fix: resolve node error by using STRING for labels and improve model loader diagnostics
Hawk Lee
2025-12-31 10:38:56 +08:00
Hawk Lee