ComfyUI-Frame-Interpolation

ComfyUI Frame Interpolation (ComfyUI VFI) (WIP)

A custom node set for Video Frame Interpolation in ComfyUI. UPDATE Memory management is improved. Now this extension takes less RAM and VRAM than before.

UPDATE 2 VFI nodes now accept scheduling multipiler values

Nodes

• KSampler Gradually Adding More Denoise (efficient)
• GMFSS Fortuna VFI
• IFRNet VFI
• IFUnet VFI
• M2M VFI
• RIFE VFI (4.0 - 4.9) (Note that option fast_mode won't do anything from v4.5+ as contextnet is removed)
• FILM VFI
• Sepconv VFI
• AMT VFI
• Make Interpolation State List
• STMFNet VFI (requires at least 4 frames, can only do 2x interpolation for now)
• FLAVR VFI (same conditions as STMFNet)
• ATM-VFI (only supports 2x interpolation)
• MoMo VFI (only supports 2x interpolation)

Install

ComfyUI Manager

Incompatibile issue with it is now fixed

Following this guide to install this extension

https://github.com/ltdrdata/ComfyUI-Manager#how-to-use

Command-line

Windows

Run install.bat

For Window users, if you are having trouble with cupy, please run install.bat instead of install-cupy.py or python install.py.

Linux

Open your shell app and start venv if it is used for ComfyUI. Run:

python install.py

Support for non-CUDA device (experimental)

If you don't have a NVidia card, you can try taichi ops backend powered by Taichi Lang

On Windows, you can install it by running install.bat or pip install taichi on Linux

Then change value of ops_backend from cupy to taichi in config.yaml

If NotImplementedError appears, a VFI node in the workflow isn't supported by taichi

Usage

All VFI nodes can be accessed in category ComfyUI-Frame-Interpolation/VFI if the installation is successful and require a IMAGE containing frames (at least 2, or at least 4 for STMF-Net/FLAVR).

Regarding STMFNet and FLAVR, if you only have two or three frames, you should use: Load Images -> Other VFI node (FILM is recommended in this case) with multiplier=4 -> STMFNet VFI/FLAVR VFI

clear_cache_after_n_frames is used to avoid out-of-memory. Decreasing it makes the chance lower but also increases processing time.

It is recommended to use LoadImages (LoadImagesFromDirectory) from ComfyUI-Advanced-ControlNet and ComfyUI-VideoHelperSuite along side with this extension.

Example

Simple workflow

Workflow metadata isn't embeded Download these two images anime0.png and anime1.png and put them into a folder like E:\test in this image.

Complex workflow

It's used in AnimationDiff (can load workflow metadata)

Credit

Big thanks for styler00dollar for making VSGAN-tensorrt-docker. About 99% the code of this repo comes from it.

Citation for each VFI node:

GMFSS Fortuna

The All-In-One GMFSS: Dedicated for Anime Video Frame Interpolation

https://github.com/98mxr/GMFSS_Fortuna

IFRNet

@InProceedings{Kong_2022_CVPR, 
  author = {Kong, Lingtong and Jiang, Boyuan and Luo, Donghao and Chu, Wenqing and Huang, Xiaoming and Tai, Ying and Wang, Chengjie and Yang, Jie}, 
  title = {IFRNet: Intermediate Feature Refine Network for Efficient Frame Interpolation}, 
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, 
  year = {2022}
}

IFUnet

RIFE with IFUNet, FusionNet and RefineNet

https://github.com/98mxr/IFUNet

M2M

@InProceedings{hu2022m2m,
    title={Many-to-many Splatting for Efficient Video Frame Interpolation},
    author={Hu, Ping and Niklaus, Simon and Sclaroff, Stan and Saenko, Kate},
    journal={CVPR},
    year={2022}
    }

RIFE