waifu2x-ncnn-vulkan
AI-powered image upscaling with Vulkan
TLDR
SYNOPSIS
waifu2x-ncnn-vulkan [-i input] [-o output] [-s scale] [-n noise]
DESCRIPTION
waifu2x-ncnn-vulkan is an AI-powered image upscaling and denoising tool that uses convolutional neural networks to enlarge images while preserving detail. It is particularly effective for anime-style artwork and illustrations, producing cleaner results than traditional interpolation methods.
The tool supports 2x upscaling and multiple noise reduction levels for removing JPEG compression artifacts. It uses the ncnn neural network inference framework with Vulkan GPU acceleration, enabling fast processing on a wide range of graphics hardware without requiring CUDA or specific vendor drivers.
Processing can be tuned with tile size options to manage GPU memory usage for large images, and specific GPUs can be selected on multi-GPU systems. Input and output support common image formats including PNG, JPEG, and WebP.
PARAMETERS
-i FILE
Input file.-o FILE
Output file.-s SCALE
Scale ratio (2).-n LEVEL
Noise level (-1,0,1,2,3).-g GPU
GPU device.-t SIZE
Tile size.
CAVEATS
GPU recommended. Best for anime/illustrations. Large VRAM for big images.
HISTORY
waifu2x-ncnn-vulkan is a Vulkan implementation of waifu2x, an AI image upscaler originally by nagadomi.
SEE ALSO
imagemagick(1), realsr-ncnn-vulkan(1)

