A simple tool for video manipulations, such as encoding, downscaling, upscaling, frame interpolation, ...
The main goal is to make an app for video alterations using AI with .onnx models (such as RealCUGAN, RealESRGAN, Waifu2x, RIFE, ...) and Nvidia's acceleration technologies (CUDA/TensorRT) in a modular pipelining approach. It would also encode the final result using ffmpeg with AV1 codec.
- ✅ FINISHING THE APP
- ⬛ Non-Nvidia GPU support
- ⬛ Linux support (should be natif, but NixOS have some problems with dependencies)
- an Nvidia gpu
- ffmpeg installed and added to the PATH
- Go to the releases
- Download
- Extract with all the files in the SAME folder
- Enjoy
- RealESRGAN
- RealCUGAN
- RIFE
- Bilinear
- Bicubic
- Lanczos
- InterArea (from cv2 through opencv-python)
- flet: a simple framework for building Flutter apps in Python
- PyTorch: a widely used Python deep learning library
- numpy: a well-known Python library for multi-dimensionnal arrays/matrices computing
- ONNX runtime: a performance-focused AI engine for
.onnxmodels - PyAV: a fast Python FFmpeg binding
All AI models are from this beautiful repository. Thanks to AmusementClub !
This project is under MIT License, see LICENSE for more informations.
Note
If you find this project useful, please consider giving it a star!