Ryan Holanda
9ebd2a85e5
- AudioToTensor: Converts ComfyUI AUDIO to VCAUDIOTENSOR (mono, resampled) - TensorToAudio: Converts VCAUDIOTENSOR back to ComfyUI AUDIO - Compatible with MuseTalk-KJ, VoiceCraft, and any Whisper-based node - Zero additional dependencies (uses PyTorch + torchaudio)
1.2 KiB
1.2 KiB
ComfyUI-AudioBridge
Bridges ComfyUI's native AUDIO type to raw audio tensors (VCAUDIOTENSOR) used by nodes like MuseTalk-KJ and VoiceCraft.
Nodes
| Node | Description |
|---|---|
| Audio → Tensor | Converts AUDIO → VCAUDIOTENSOR (mono, resampled to target rate) |
| Tensor → Audio | Converts VCAUDIOTENSOR → AUDIO (wraps tensor back for standard nodes) |
Why?
ComfyUI's native AUDIO type is a dict ({waveform, sample_rate}), but some custom nodes expect VHS_AUDIO (a callable) or VCAUDIOTENSOR (a raw torch tensor). This node bridges the gap without requiring VideoHelperSuite or duplicate audio loading.
Install
cd ComfyUI/custom_nodes
git clone https://github.com/RyanHolanda/ComfyUI-AudioBridge.git
No additional dependencies required — uses PyTorch and torchaudio (already included with ComfyUI).
Usage Example
LoadAudio → [Audio → Tensor] → whisper_to_features → MuseTalk Sampler