waifu2x
Open-source AI upscaler for anime-style art, MIT licensed.
About
waifu2x is an open-source image super-resolution and noise-reduction program created by nagadomi, inspired by the SRCNN research paper. It uses deep convolutional neural networks specifically tuned for anime-style art and illustrations, upscaling images up to 2x with noise reduction while preserving linework and color transitions better than photo-trained upscalers.
The original implementation is written in Lua with CUDA support and released under the MIT License, and the ecosystem includes GPU-accelerated variants like waifu2x-ncnn-vulkan for Windows, macOS, and Linux, plus free browser-based mirrors requiring no installation. It remains a staple tool for anime communities and illustrators despite less active development on the original repo.
What it can do
- 2x AI upscaling tuned for anime-style art
- Adjustable noise reduction levels
- Batch processing of image folders
- MIT-licensed with free web mirrors
Setup
Use a free web mirror, or clone the GitHub repo and run locally with CUDA/Vulkan.
Prerequisites
For local use: Lua/Torch or a Vulkan-capable GPU for accelerated builds.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 89; Trending score 74/100 from researcher assessment.
★ 27,500 GitHub stars
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.