R-ESRGAN-AnimeVideo-UI (reav-ui)

Upscale your videos using the R-ESRGAN AnimeVideo model

Check it out here on my Github!👋

📝 Usage

• Run the portable executable file or the R-ESRGAN-AnimeVideo-UI.pyw script to launch the UI.

• Wait for the required files to downloaded. ~210mb - (ffmpeg, ffprobe realesrgan-ncnn-vulkan-20220424-windows, models from the repo)

  • The files will download in the background, and the app will open when all files are downloaded.

• Select a video file that you want to upscale.

• Follow the steps labeled in the UI.

Note: The Upscale and Merge operations delete the previous frames by default. If you want to keep the frames, make sure to enable the Keep Frames option.

After Extracting or Upscaling, you have the option to scale the frames to any size, either by percent or exact resolution.

mp4, gif, avi, mkv, webm, mov, m4v, wmv, are supported.

• Other Uses:

  • Batch Upscale a folder of images.

  • Upscale a single image.

  • You can also use these models for upscaling: RealESRGAN_General_x4_3, realesrgan-x4plus, realesrgan-x4plus-anime

  • Output any video as a gif, and/or output gif as mp4.

🤷 Why?

Why use this over an application like Chainner, or Topaz Video AI?

While both applications are undoubtedly more than capable, (among others, I'm sure) I found xinntao's implementation is extremely fast compared to Chainner. And of course free is always better than Topaz Video AI's high cost.

Other upscale models are popular, but realesr-animevideov3 remains a favorite of mine for multiple reasons.

• It's very fast. Speed Tests⏱️

• While it does have somewhat heavy smoothing, it doesn't totally destroy fine details like grain.

• In my opinion, it has superb line reconstruction compared to others like RealESRGAN_x4plus_anime_6B

Check out these comparisons.✨

🚩 Requirements

You don't need to worry about anything if you're using the portable/executable version.

You must have Python 3.10+ installed to the system PATH

Running the script will automatically fulfill all requirements.

The pillow library will be downloaded and installed (if not already available) upon first launch.

ffmpeg and ffprobe will be downloaded upon first launch.

realesrgan-ncnn-vulkan-20220424-windows.zip and the required models will be downloaded upon launch.

📜 Version History

Check out the releases page for all changelogs.

👥 Credits

ffmpeg-6.0-essentials: https://ffmpeg.org/

Real-ESRGAN_portable: https://github.com/xinntao/Real-ESRGAN#portable-executable-files-ncnn

Take a look at my others articles!

This page will not be updated as frequently as the github page, some info may be out of date.