Upscayl

upscayl/upscaylReviewed by AI

What it does

Upscayl is a desktop program that makes a low-resolution picture bigger while filling in detail, using AI models that run on your own graphics card. You pick an image or a whole folder, choose a model and a scale, and it saves the enlarged copy; nothing is uploaded and there is no account. It is for anyone with old photos, small scans or pixelated images. The AI models come inside the download, so there is nothing else to fetch, but you do need a graphics card that supports Vulkan, and the project warns that many built-in laptop graphics chips will not work.

How easy it is to install

A · download and open · Windows

Windows A
macOS A
Linux A

Every system has a ready download: a Windows installer (upscayl-2.15.0-win.exe, 249 MB), a Mac .dmg you drag to Applications, and for Linux an AppImage, .deb and .rpm plus Flathub, Snap and AUR packages. The models are bundled, so there is nothing to set up after installing — but the program only runs the AI if your graphics card supports Vulkan.

Numbers on GitHub

Stars: 51,524on Oct 31, 2026
Downloads: 84,110,417279,426 this week
Growth: +419stars in 5 days

Stars over time

Facts

Runs on: Windows, macOS, Linux
You need to install: Graphics card
To build from source: Node.js
Disk space: 249 MB–746 MB · from the installer
Accounts or keys: None
Cost: Free · Free and open source (AGPL-3.0). The download includes the AI models; the project asks only for donations.
License: AGPL-3.0 (Copyleft)
Can replace: Topaz Gigapixel AI
Last update: Mar 27, 2026

Risk signals (5)

Attentionneeds vulkan graphics card
README, marked IMPORTANT: "You'll need a Vulkan compatible GPU (Graphics Card) to upscale images. Many iGPUs (integrated graphics) do not work but, no harm in trying :)", and the FAQ repeats "Yes, unfortunately. NCNN Vulkan requires a Vulkan-compatible GPU. Upscayl won't work with most iGPUs or CPUs." It can install cleanly and still refuse to upscale.
AI review
Attentionno new release in almost two years
The newest release is v2.15.0 from 2024-12-25 and the last commit on the analyzed branch is from 2026-03-27, 216 days before this check. The README still announces "v2.15 is out!".
AI review
Attentioninstaller triggers os warning
The Windows instructions say: "If you get a SmartScreen warning - click 'More Info' and then 'Run Anyway' OR press 'YES' on the unverified publisher dialog", and the Mac steps tell you to right-click the app and press 'Open' twice. package.json sets `"dmg": { "sign": "false" }`. You are being asked to step past your system's own check.
AI review
Infodeclared alternative comes from a sponsor banner
The checks picked up a declared alternative to Postman from the line "Requestly - Free & Open-Source alternative to Postman". That line is a paid sponsor banner at the top of the README (alongside one for Warp), not a claim about Upscayl, which only handles images.
AI review
Infoextra models from elsewhere
README points to a separate repository for more models ("More models can be taken from here: https://github.com/upscayl/custom-models") and credits community models hosted on openmodeldb.info. Anything you add yourself comes from outside this download.
AI review

Evidence is shown in its original language.

What Radar could not check

Whether your particular graphics card is one of the supported ones, which decides if the program works at all
How good the result looks on your own pictures, and how long each image takes
Whether a build from December 2024 still runs on the newest Windows and macOS versions
Real disk use once you add extra community models
What the separate macOS App Store listing costs or contains, which Radar did not check