main
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

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 toweak. The config is located atuser/__manager/config.ini
- If you get the error
- Restart ComfyUI.
Manual Install
- Clone this repository into the
custom_nodesdirectory:
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.
Languages
Python
100%