QualityScaler

Why we included this project

QualityScaler is a Windows application that cleans up and enlarges low-resolution, compressed, or noisy images and video using established super-resolution models such as Real-ESRGAN, BSRGAN, and IRCNN. It wraps those models in a Python GUI built on ONNX Runtime DirectML, so it runs on AMD, Intel, and Nvidia GPUs without you having to set up a deep learning environment. The app handles both stills and video and includes practical touches like automatic tiling to manage VRAM limits, multi-GPU processing, batch upscaling, resizing before upscaling, blending between original and enhanced output, and pause/resume for long jobs. Everything runs locally, so no internet connection is needed, and it reads a wide range of common formats. For photographers, video editors, and anime or manga fans who want AI restoration without scripting, this is a solid all-in-one option.