← See All Custom Node Packs

ComfyUI-Frame-Interpolation

1040
Updated 3 months ago
View on GitHub →See Common Issues →

Run ComfyUI Easily with InstaSD

Skip the complex setup. InstaSD helps creative professionals build workflows and deploy them to the world:

  • One-click deployment
  • Any model, any node
  • Powerful GPUs for rapid iteration
Get Started

Available Nodes

Documentation

ComfyUI Frame Interpolation Repository

Overview

The ComfyUI Frame Interpolation repository, also known as ComfyUI VFI, is a work-in-progress custom node set designed for Video Frame Interpolation within the ComfyUI environment. It aims to enhance video processing capabilities by providing several nodes specifically tailored for frame interpolation tasks.

Installation Guide

Through ComfyUI Manager

Previously reported compatibility issues with ComfyUI Manager have been resolved. To install this extension, follow the installation guide available at ComfyUI Manager's How To Use.

Command-line Installation

Windows

For Windows users, execute the install.bat script. If you encounter issues with cupy, ensure you run install.bat instead of install-cupy.py or python install.py.

Linux

For Linux users, activate your virtual environment (if used for ComfyUI), then run:

python install.py

Support for Non-CUDA Devices (Experimental)

If you lack an NVIDIA card, you can utilize the taichi ops backend powered by Taichi Lang.

  • On Windows, install it by running install.bat.
  • On Linux, execute pip install taichi.

Subsequently, change the ops_backend setting from cupy to taichi in config.yaml. Note that if a NotImplementedError occurs, it indicates a VFI node in the workflow is unsupported by taichi.

Repository Purpose

This repository is dedicated to video frame interpolation tasks using ComfyUI. It provides custom nodes that facilitate the interpolation of frames for smoother video sequences and transitions.

Provided Nodes

The repository includes various nodes optimized for video frame interpolation tasks:

Special Features

  • Improved Memory Management: Recent updates have enhanced memory management, resulting in reduced RAM and VRAM usage.
  • Scheduling Multiplier Support: VFI nodes now support scheduling multiplier values.
  • Non-CUDA Device Support: Experimental support for non-CUDA devices via the taichi backend.

Usage in ComfyUI Workflows

All VFI nodes are accessible in the category ComfyUI-Frame-Interpolation/VFI upon successful installation. They require an IMAGE containing multiple frames (minimum two, or four for STMFNet/FLAVR). To mitigate out-of-memory issues, adjust clear_cache_after_n_frames by decreasing its value to reduce memory load while potentially increasing processing time.

For enhanced functionality, it is recommended to use LoadImages from ComfyUI-Advanced-ControlNet in combination with ComfyUI-VideoHelperSuite.

Example Workflows

  • Simple Workflow: Utilize two images and configure the nodes to perform frame interpolation.
  • Complex Workflow: Designed for more advanced animations and leveraging comprehensive ComfyUI features.

Credit and Additional Resources

The repository is heavily based on code from VSGAN-tensorrt-docker, with significant contributions to the frame interpolation functionality. Detailed references and citations for each VFI node are provided in the repository documentation to acknowledge the contributions from various research efforts.