Sooner or later we all need a larger image than usual, even if we have to invent pixels in the process. But thanks to new algorithms, upscaling errors and artifacts are becoming less frequent. Moreover, many free and open-source tools continue to improve their interfaces. One of the most interesting is "Waifu2x-ncnn-Vulkan GUI Edition Reloaded", which offers a complete graphical environment for waifu2x-ncnn-vulkan and eliminates the need for long console commands.
In the past we have talked about waifu2x-caffe, which has the advantages of offline editing and CUDA support for Nvidia graphics cards. But variations, optimized algorithms and alternatives have slowly emerged that can deliver better results under more flexible hardware configurations.
A good example is waifu2x-ncnn-vulkan, which works on anything with Vulkan support (Intel, AMD, Nvidia, and even Apple's new silicon). However, its command-line execution is not very comfortable and it urgently needs an interface. Fortunately, XyLe-GBP on GitHub took the trouble to create "Waifu2x-ncnn-Vulkan GUI Edition Reloaded", a fast and easy-to-use graphical environment.
How to enlarge images without losing quality with "Waifu2x-ncnn-Vulkan GUI Edition Reloaded"
All you need to do to get started is download one of the available versions (standard or portable installation) and run its executable. Then drag the image or video onto the interface and go to the corresponding upscaling section in the Config menu. Our tests focus on images, but the general parameters are the same.
The program provides access to three different engines: Waifu2x, Real-CUGAN, and Real-ESRGAN. Depending on the engine chosen, the General and Advanced tabs allow selection and modification of various values.
For example, Real-ESRGAN has three optimized models (realesrgan-x4plus, realesrnet-x4plus, and realesrgan-x4plus-anime), but its minimum upscaling is 4x, while Waifu2x and Real-CUGAN accept 2x or 3x.
There is plenty of room to experiment here, but even under its normal factory settings, "Waifu2x-ncnn-Vulkan GUI Edition Reloaded" produces very good results, especially if we feed it images from Stable Diffusion. The version we installed in our tutorial can upscale images on its own with the requirement of more VRAM on the graphics card (more than 6 GB), therefore, I recommend using "Waifu2x-ncnn-Vulkan GUI Edition Reloaded" independently.
Official site and download: Click here
