R
Real-ESRGAN
Description
Real-ESRGAN is an open set of algorithms for restoring and upscaling images and video. It extends ESRGAN and, according to the authors, is trained on purely synthetic data. The repository has a script for video, a separate model for anime video, and optional face enhancement through GFPGAN. Ready-made portable programs for Windows, Linux and macOS run on Intel, AMD and Nvidia GPUs and need neither CUDA nor PyTorch; the Python version is also available. The code is published under the BSD-3-Clause license. The authors do not state GPU memory requirements, and the last tagged release on the repository dates from 2022.
Pros and cons
Pros
- BSD-3-Clause license, commercial use allowed
- Ready portable programs for Windows, Linux and macOS, no CUDA or PyTorch needed
- Separate models for anime video
- Optional face enhancement
Cons
- Last tagged release dates from 2022
- GPU memory requirements are not stated by the authors
- Video is processed with a command-line script, no ready video editor
- Trained on synthetic data, so fine detail can be invented
Key facts
- License
- BSD-3-Clause
- Commercial use
- Allowed
- Developer
- Real-ESRGAN authors
- Checked on
- Oct 21, 2026
Editorial score
How strongly the AI Expertise editors recommend this open model for each of its tasks right now, on a 0–10 scale. The score combines output quality, capabilities, ease of launch, license openness and freshness.