type-two
|
110b2298b7
|
audio/video/text operator suite: stt_local, lipsync_local, voice_clone_local, wan_video, llm_local
- stt_local: Whisper large-v3-turbo MLX, transcript + timestamped json
- lipsync_local: Rhubarb visemes (CPU lane), dialog-guided
- voice_clone_local: Chatterbox on MPS (needs setuptools<81 for perth/pkg_resources)
- wan_video: Wan 2.2 TI2V-5B t2v/i2v via resident ComfyUI, frames->mp4
- llm_local: Qwen3-30B-A3B-4bit dialogue writer (primary-only)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
2026-07-17 18:26:51 +10:00 |
|