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