Our take
- Activity
- 19
- Community
- 59
- Issues
- 54
- Pull requests
- 20
+1,791 stars in the last 30 days
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
52,295Forks
2,738Change since Nov 11, 2026: +100 stars·+1 forks
Overview
Upscayl is a free desktop image upscaler that uses AI models to reconstruct details when enlarging low-resolution or pixelated images. It offers a graphical interface for applying models based on Real-ESRGAN, with processing powered by Vulkan.
It is useful when you want to enlarge an image while improving its apparent detail, rather than simply scaling pixels. It is not designed to fix out-of-focus or heavily blurred photos.
Key Features
- Enlarge and enhance images using AI models.
- Provides a desktop app for Linux, macOS, and Windows.
- Uses Vulkan through its NCNN-based backend.
- Supports multiple image models, with additional models available from the linked custom-models project.
- Includes batch processing and image post-processing options, as described in the README.
- Offers a separate command-line tool, upscayl-ncnn.
Use Cases
- Photographers can enlarge small or pixelated images when a larger output is needed.
- Designers and illustrators can improve the apparent detail of low-resolution assets before using them in larger layouts.
- Users restoring older digital images can try AI upscaling when the source is low-resolution but still reasonably in focus.
- Developers or automation-minded users can use the separate command-line backend for image-upscaling workflows.
Getting Started
Download a release for your operating system from GitHub Releases or visit the official website. A Vulkan-compatible GPU is required. See the README for installation instructions and platform-specific details.
Considerations
- Upscaling requires a Vulkan-compatible GPU. The README notes that many integrated GPUs and CPUs will not work.
- AI upscaling estimates image details. It cannot reliably correct blur or out-of-focus images.
- The app is licensed under AGPL-3.0, which has obligations for distribution and modified network-accessible software.
- The repository is active and not archived, but project activity and compatibility may vary across operating systems and GPU hardware.
Found this useful?
Share it with someone who would like upscayl.