gpu-font matches a screenshot to the closest fonts from multiple open catalogs directly in the browser.
- for
- Designers and developers who need to identify fonts from images without uploading data.
- pricing
- open source
- license
- MIT
- v0.2.00.2.018 days ago
Key features
- WebGPU acceleration — Runs on WebGPU for ~0.06 s inference, with CPU fallback.
- Multiple catalogs — Searches Google Fonts, DaFont, Adobe Fonts, GitHub, Debian and more.
- Local processing — All computation happens in the browser; no image is sent to a server.
- Custom font indexing — Add your own installed or folder fonts to the catalog for personal matching.
- Weight & italic detection — Each match includes the closest face’s weight and italic style.
- Node.js API — Install via npm and use the matching engine from server-side code.
Use cases
- Identify a web font used in a screenshot of a landing page.
- Find open-source alternatives to a commercial font for a redesign.
- Quickly verify the font family of a logo or UI element during development.
gpu-font FAQ
Is any data uploaded to a server?+
No, the model runs entirely in the browser or Node, so nothing is sent to a remote server.
Can it detect commercial fonts?+
It only indexes openly licensed fonts; commercial fonts are matched to the closest open look-alikes unless you index your own copy.
Does it work with non-Latin scripts?+
Yes, but with lower accuracy; Cyrillic, Arabic and Chinese are supported to varying degrees.
Can I add my own fonts to the search catalog?+
Yes, you can index local fonts or folders; they stay in your browser and are searchable.
What are the hardware requirements?+
A browser with WebGPU support for fast inference; otherwise it falls back to CPU with slower performance.
Is the project open source?+
The code is MIT-licensed and available on GitHub, with the model and training pipeline also provided.
Summarized by DevHunt from dy.github.io · Sep 28, 2026. Details may change; check the official site.
