Skip to content
buildbay.
Upscayl logo

Upscayl

upscayl · Images

GitHub stars

50,038
—Collecting 30d data

Stats updated

49,941 stars on Sep 27 to 50,038 stars on Oct 1, up 97.5 measured star snapshots

What is Upscayl?

Upscayl provides a dedicated desktop workflow for enlarging low-resolution or pixelated images with local AI models. It packages the open NCNN/Vulkan backend for common desktop operating systems, offers several enhancement models and output controls, and supports both individual images and batches without requiring a hosted processing account.

Upscayl is a free, open-source desktop app for upscaling low-resolution images with AI models. It runs on Linux, macOS, and Windows, and can batch upscale images.

Core capabilities

  • Enlarge and enhance low-resolution images with local AI upscaling models
  • Process individual files or image batches on Linux, macOS, and Windows
  • Choose bundled or custom models and control output scale and format

A focused desktop upscaling workflow

Upscayl packages model selection, scale, output location, and before-and-after review in a desktop interface. Linux, macOS, and Windows builds use the open upscayl-ncnn backend, keeping the default enhancement process on a Vulkan-compatible local GPU.

  • desktop
  • Vulkan
  • NCNN
  • local

Models, batches, and clear limits

Users can process a single image or a batch, switch among supplied models, and add compatible custom models. Official guidance emphasizes low-resolution or pixelated sources rather than fundamentally out-of-focus images. The usual desktop path needs a Vulkan-compatible GPU; CPU-only workarounds are platform-dependent and may be slow.

  • batch
  • custom models
  • GPU
  • image enhancement

Where it fits

Use cases

  1. 01

    Prepare a small image for a larger layout

    Select a low-resolution or pixelated source, choose an upscaling model and target scale, compare the result, and export a larger file for presentation, publishing, or archival use.

  2. 02

    Upscale a folder of related assets

    Run batch processing across a group of source images with consistent model and output settings, then let Upscayl finish enhancement before it applies supported post-processing steps.

  3. 03

    Test a custom enhancement model

    Add compatible custom models or convert a model for the NCNN backend, select it from the desktop application, and compare its output against the bundled choices on representative images.