ComfyUI Extension: comfyui_momo

Authored by chameleon-ai

Created

Updated

0 stars

Implementation of MoMo interpolation for ComfyUI

Custom Nodes (0)

    README

    ComfyUI MoMo

    Implementation of MoMo video frame interpolation for ComfyUI. This does simple 2x interpolation.

    The original MoMo code was written by Jaihyun Lew. This is a ComfyUI node implementation with minimal changes.

    Disentangled Motion Modeling for Video Frame Interpolation, AAAI 2025 momo cover

    Installation

    ComfUI Manager

    • Install ComfyUI-Manager
    • Look up this extension in ComfyUI-Manager or install via URL
      • If you get the error This action is not allowed with this security level configuration, you will have to edit the security policy to weak. The config is located at user/__manager/config.ini
    • Restart ComfyUI.

    Manual Install

    • Clone this repository into the custom_nodes directory:
    cd custom_nodes
git clone https://github.com/chameleon-ai/comfyui_momo.git
    • Activate the ComfyUI virtual environment and install dependencies
    source venv/bin/activate
pip install -r custom_nodes/comfyui_momo/requirements.txt

    Usage

    On first run, the model will be downloaded to models/diffusers/momo_full.pth
    You can manually download the model here

    The node is located under image/animation/MoMo Video Frame Interpolation
    An example workflow is included.

    Dependencies:

    torch
accelerate
diffusers
einops

    Tested in python 3.13 on Linux+AMD. The original MoMo implementation was for python 3.10, so it should work for that and everything in between.