Bringing 120+ Audio AI Models to ComfyUI in One Custom Node
I created ComfyUI-AudioCPP about two months ago, but made pretty much zero progress 😅. Today, a community contributor finally got things moving by implementing the first set of nodes!
The goal is to bring the capabilities of audio.cpp into ComfyUI:
- All-in-one: TTS, ASR, voice conversion, music generation, audio enhancement, separation, and more through a unified engine.
- Broad model coverage: 120+ audio model families, without maintaining separate inference implementations for each.
- Unified interface: Many models share the same interface in audio.cpp, so supporting additional models in ComfyUI can come almost for free once the corresponding task nodes are implemented.
- High performance: Optimized native C++/GGML inference across CUDA, Vulkan, Metal, CPU, and AMD (experimental), without Python inference dependencies.
The initial contribution provides TTS nodes, including voice cloning support. It's still early, and there's plenty of room to expand into other audio tasks and improve the ComfyUI integration.
Contributions are very welcome! Whether you want to add nodes, improve workflows, test different models/backends, or help polish the integration, feel free to open a PR!
Repo: 0xShug0/ComfyUI-AudioCPP
19 points · 2 comments
Mentions
Bringing 120+ Audio AI Models to ComfyUI in One Custom Node I created ComfyUI-AudioCPP about two months ago, but made pretty much zero progress 😅. Today, a community contributor finally got things moving by implementing the first set of nodes! The goal is to bring the capabilities of audio.cpp into ComfyUI: - All-in-one: TTS, ASR, voice conversion, music generation, audio enhancement, separation, and more through a unified engine. - Broad model coverage: 120+ audio model families, without maintaining separate inference implementations for each. - Unified interface: Many models share the same interface in audio.cpp, so supporting additional models in ComfyUI can come almost for free once the corresponding task nodes are implemented. - High performance: Optimized native C++/GGML inference across CUDA, Vulkan, Metal, CPU, and AMD (experimental), without Python inference dependencies. The initial contribution provides TTS nodes, including voice cloning support. It's still early, and there's plenty of room to expand into other audio tasks and improve the ComfyUI integration. Contributions are very welcome! Whether you want to add nodes, improve workflows, test different models/backends, or help polish the integration, feel free to open a PR! Repo: 0xShug0/ComfyUI-AudioCPP