Open Source Linux Image Upscalers

Free and Open Source AI Image Upscaler for Linux, MacOS and Windows built with Linux-First philosophy. Upscayl is a cross-platform application built with the Linux-first philosophy. This means that we prioritize Linux builds over others but that doesn't mean we'll break things for other OSes. Upscayl does not work without a GPU, sorry. You'll need a Vulkan-compatible GPU to upscale images. CPU or iGPU won't work. You can also download the flatpak version and double-click the flatpak file to install via Store but wait for the full release, we'll be pushing it to Flathub for easy access. Upscayl uses AI models to enhance your images by guessing what the details could be. It uses Real-ESRGAN (and more in the future) model to achieve this. The CLI tool is called real-esrgan-ncnn-vulkan and it's available on the Real-ESRGAN repository.

Qualityscaler is a Windows app that uses BSRGAN Artificial Intelligence to enhance, enlarge and reduce noise in photographs and videos. QualityScaler is completely written in Python, from the backend to the front end. Image/list of images upscale. Video upscale. Drag&drop files [image / multiple images/video] Automatic image tiling and merging to avoid gpu VRAM limitation. Resize image/video before upscaling. Multiple Gpu support. Compatible images - png, jpeg, bmp, webp, tif. Compatible video - mp4, wemb, gif, mkv, flv, avi, mov, qt.

Create textures, concept art, background assets, and more with a simple text prompt. Use the 'Seamless' option to create textures that tile perfectly with no visible seam. Texture entire scenes with 'Project Dream Texture' and depth to image. Re-style animations with the Cycles render pass. Run the models on your machine to iterate without slowdowns from a service. Create textures, concept art, and more with text prompts. Learn how to use the various configuration options to get exactly what you're looking for. Texture entire models and scenes with depth to image. Inpaint to fix up images and convert existing textures into seamless ones automatically. Outpaint to increase the size of an image by extending it in any direction. Perform style transfer and create novel animations with Stable Diffusion as a post processing step. Dream Textures has been tested with CUDA and Apple Silicon GPUs. Over 4GB of VRAM is recommended.

Clarity AI Upscaler is an open-source AI image enhancement tool designed to increase the resolution and visual quality of images using modern generative techniques. The system uses deep learning models based on diffusion and other image generation methods to reconstruct high-resolution versions of low-resolution images while preserving important visual details. Unlike traditional interpolation-based upscaling algorithms, the system generates additional visual information that improves perceived clarity and sharpness. The project is intended as a free and open alternative to commercial AI upscaling tools, allowing developers and digital artists to run the technology locally or integrate it into their own workflows. The repository includes a full application environment with scripts, configuration files, and model support that allow users to run the upscaler as a standalone tool or integrate it into other pipelines.

This is an ESRGAN model trained specifically for upscaling GameCube and Wii textures, but it can of course be used for other textures from that period, like PlayStation 2, Xbox or PC games from that time. It can be used for all image formats supported by Gamecube and Wii hardware and can remove its typical artifacts like CMPR Block Compression (DXT1 algorithm, also known as BC1), color palette errors, color reduction up to 8bit color depth and 1bit alpha depth. I recommend chaiNNer, which offers a graphical user interface and is based on a node system that can be used to solve very simple as well as very complex tasks effectively by "chaining" nodes together. You get the best quality if you upscale the alpha channel separately.

Single-Image Super-Resolution for Anime-Style Art using Deep Convolutional Neural Networks. And it supports photo. You can train your own model, change image size, reduce image noise, upscale and customize your image's style. It provides the option of converting and downloading your edited images.

This is a GUI app that can upscale anime-styled images, gifs, and videos using waifu2x or Real-ESRGAN Anime. It aims to be simple to use, but you can access more advanced settings in the top bar. Upscaling videos and gifs will take a very long time without a high-end graphics card! The Real-ESRGAN Anime upscaler is much faster and gives better results than waifu2x, so it is recommended to use it instead. Real-ESRGAN only supports scale factors between 2-4, and all other waifu2x options are ignored. By setting the scale factor to 4x, it will use the slower Anime4x model but this gives better results, so it is recommended to upscale images at 4x when using Real-ESRGAN. Apply effects such as speed or reverse (animated images/videos). Customize settings (noise, scale, mode, framerate, etc.). Upscale multiple images (or multiple frames of an animated image/video) concurrently. Change the upscaler (waifu2x or Real-ESRGAN Anime).

To speed up, use "-b 256 -j 4" options. The original is there: https://github.com/WL-Amigo/waifu2x-converter-cpp Here are some minor cosmetic changes (switch to control memory usage and slightly reformatted output) and files needed to build under MinGW32-W64. To recompile (in Linux), run "make clean; make". 2016-10-17 UPDATE: added simple GUI. 2016-10-18 FIX: corrected typo, added some crashes handling. 2016-11-20 UPDATE: - added "TTA" and "number of passes" options; - enabled RGB models (without internal upscaling); - updated JPEG, TIFF, WebP and OpenEXR libraries; - modified OpenCV to enable JPEG compression in TIFF images (sources included); - replaced models with Nagadomi's ones (https://github.com/nagadomi/waifu2x). 2016-12-28: Happy New Year and Merry Christmas! - "Stop" button works; updated OpenCV, WebP and JasPer libraries; - fix: didn't filter images when scale < 0.5; fixed conversion to 8 bpp in OpenCV and reading of some JPEG-compressed TIFFs in LibTIFF.

The goal of this project is to upscale and improve the quality of low-resolution images. This project contains Keras implementations of different Residual Dense Networks for Single Image Super-Resolution (ISR) as well as scripts to train these networks using content and adversarial loss components. Docker scripts and Google Colab notebooks are available to carry training and prediction. Also, we provide scripts to facilitate training on the cloud with AWS and Nvidia-docker with only a few commands. When training your own model, start with only PSNR loss (50+ epochs, depending on the dataset) and only then introduce GANS and feature loss. This can be controlled by the loss weights argument. The weights used to produce these images are available directly when creating the model object. ISR is compatible with Python 3.6 and is distributed under the Apache 2.0 license.

The original is here: https://github.com/nagadomi/waifu2x Differences: - CPU only, no CUDA needed; - double-precision version; - GPU and CPU-trained models (on the fly conversion in memory); - adapted for Lua v5.2 (works with Torch 7 on x32 Ubuntu). 31-12-2016: - removed unneeded data copies, left from CUDA processing; - removed large duplicate files (see "ReadMe" before use).