IMTalker ComfyUI

A ComfyUI plugin for IMTalker, an audio-driven talking face generation model that creates realistic facial animations from audio input.

Installation

In the ./ComfyUI/custom_nodes directory, run the following:

git clone https://github.com/benda1989/imtalker-comfyui.git
cd imtalker-comfyui
pip install -r requirements.txt
conda install -c conda-forge ffmpeg 

Model Setup

Download the required model checkpoints and place them in the ComfyUI/models directory:

ComfyUI/models/
├── checkpoints/
│   ├── renderer.ckpt
│   ├── generator.ckpt
│   └── wav2vec2-base-960h/
│       ├── config.json
│       ├── pytorch_model.bin
│       ├── preprocessor_config.json
│       └── feature_extractor_config.json

The models will be automatically downloaded from Hugging Face on first use.

Example Workflows

Audio-Driven Generation

Video-Driven Generation

Other Requirements

• ComfyUI: Latest version
• ComfyUI-VideoHelperSuite: For video loading/saving
• CUDA: Recommended for GPU acceleration
• FFmpeg: Required for video processing