RealScaler

Windows AI upscaler for images and video using Real-ESRGAN, fully offline

Open Source Alternative to
Repository activity
  • Stars409
  • Forks29
  • Open Issues15
License

MIT

Languages
  • Python
Get it:GitHub
RealScaler screenshot

About RealScaler

RealScaler is a Windows app that enhances, upscales, and denoises photos and video using Real-ESRGAN AI. All processing happens locally on your PC, so your source and output files never leave your machine and no internet connection is required.

It upscales both images and video, with resize-before-upscale and interpolation to blend the original with the upscaled result. Automatic tiling keeps jobs within GPU VRAM limits, multiple GPUs are supported, and video jobs can be stopped and resumed. It reads common image formats such as JPG, PNG, TIFF, BMP, WebP, and HEIC, plus video formats including MP4, MKV, AVI, MOV, and GIF.

RealScaler targets Windows 10 and 11 and needs a DirectX 12 compatible GPU with at least 4 GB of VRAM.

Key features

  • AI upscaling of images and video with Real-ESRGAN
  • Denoises photos and video
  • Automatic tiling to fit GPU VRAM limits
  • Resize before upscaling
  • Stop and resume video jobs

Details

On GitHub since
2022
Platforms
Windows
Deployment
Offline-first
Model
Real-ESRGAN
GPU
DirectX 12, 4 GB VRAM min
Media
Images and video