Nodes/comfyui-latent-astronaut-suite/ChatterBox VC Loader (Auto)
ComfyUI Node

ChatterBox VC Loader (Auto)

Download and go

By latentastronaut·Created 8 months ago·Updated 7 months ago· 2
ChatterBox VC Loader (Auto)
    • model
    deviceauto

    ChatterBox VC Loader (Auto) is the one-node, zero-file-management route to voice conversion in this pack. Give it a device choice and it downloads the official ChatterBox VC model from HuggingFace on first use, caches it, and hands you a ready model. That's the entire workflow: no .safetensors to source, no folder layout to maintain, no picking the wrong checkpoint. If the manual VC loader's file choreography isn't your thing, this is the node you want.

    How it works

    It's a thin wrapper around ChatterBox's official ChatterboxVC.from_pretrained(device=device) call. First run downloads the model from HuggingFace and caches it locally; later runs load straight from that cache. device takes auto (CUDA if present, else MPS or CPU), cuda, or cpu. As with the TTS side, the loader disables the model's watermarker afterward - the pack's author found it segfaults under ComfyUI's dependency environment - so output audio comes out clean. The result is packaged as a CHATTERBOX_VC output.

    Inputs and outputs

    • device - auto (default), cuda, or cpu

    Output: model - a CHATTERBOX_VC object for the Simple VC node. Pair this loader with ChatterBox VC (Simple), not the Manual VC node, which expects the manual loader's model object.

    Install

    Pack install via ComfyUI Manager (search "Latent Astronaut Suite") or:

    cd ComfyUI/custom_nodes
    git clone https://github.com/latentastronaut/comfyui-latent-astronaut-suite
    

    Restart. First boot's install.py installs chatterbox-tts with --no-deps (its version pins conflict with ComfyUI's stack) plus librosa; without those, the import fails. The actual model download happens on your first generation.

    Where people get burned

    Same first-run patience warning as the TTS auto-loader: the download is large and the first conversion includes model load time, so a long quiet stretch is normal. If it errors instead, restart ComfyUI once so the install script has definitely run, and check that chatterbox-tts imports. And don't try to feed this model into the Manual VC node - different object shapes, guaranteed confusion.

    Categoryaudio/latent-astronaut

    Inputs (1)

    NameTypeDefaultDescription
    deviceCOMBOautoDevice to load model on

    Outputs (1)

    NameTypeDescription
    modelCHATTERBOX_VC