GPU image upscaler and denoiser using ncnn with Vulkan on Intel, AMD, NVIDIA, and Apple Silicon
Repository activity
- Stars3.4k
- Forks238
- Open Issues80
License
MIT
Languages
- C++
- C
- CMake
Get it:Website
About waifu2x-ncnn-vulkan
waifu2x-ncnn-vulkan is a fast, GPU-accelerated version of the waifu2x image upscaler and denoiser. Built on the ncnn framework with the Vulkan API, it runs well on Intel, AMD, NVIDIA, and Apple Silicon GPUs.
It is a command-line tool that takes a single file or a whole directory as input, with adjustable noise reduction, 2x scaling, tile size, and per-stage thread counts for decoding, upscaling, and encoding. Results are written as lossless PNG or WebP.
It runs entirely on your own machine with the models bundled in the download, and is released under the MIT license.
Key features
- GPU-accelerated waifu2x upscaling and denoising
- Runs on Intel, AMD, NVIDIA, and Apple Silicon via Vulkan
- Batch processes single files or whole directories
- Adjustable noise level, scale, tile size, and thread counts
- Outputs lossless PNG or WebP
Details
- On GitHub since
- 2019
- Platforms
- Windows · macOS · Linux · CLI
- Deployment
- Offline-first
- GPU
- Intel · AMD · NVIDIA · Apple
- Framework
- ncnn · Vulkan
- License
- MIT