Real-ESRGAN

Health v2 · maintenance only

Adoption not part of the score

36593 stars · 4451 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Real-ESRGAN is a deep learning project for practical image and video restoration via super-resolution, with pretrained models for photos and anime. It ships as a Python library (pip) plus portable ncnn-Vulkan executables for Intel/AMD/Nvidia GPUs.

Use cases

• upscale low-resolution photos to 4x
• restore and denoise old or compressed images
• upscale anime images and illustrations
• enhance anime video quality frame by frame
• batch upscale images from the command line without a Python setup
• integrate super-resolution into an image processing pipeline

When to choose

• you need state-of-the-art general-purpose image upscaling
• you work with anime artwork or anime video
• you want a portable GPU executable that runs on any Vulkan-capable GPU
• you want a pip-installable Python library for restoration tasks

When to avoid

• you need real-time upscaling on weak hardware without a GPU
• you only need simple resizing without learned detail reconstruction
• you need text-focused or face-specific restoration with dedicated models
• you cannot accept GPU memory usage for large images

Facets

library · maturity stable

image-processing machine-learning deep-learning cli image-processing computer-vision machine-learning media python windows cross-platform super-resolution upscaling image-restoration denoising anime video-upscaling ncnn vulkan pytorch linux macos gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="xinntao/Real-ESRGAN")