What's That Font? AI-Powered Typography Detection

The Designer’s Dilemma

You see a beautiful typeface in a screenshot, mood board, or competitor’s site. You want to know what it is. Maybe you need it for your project. Maybe you’re just curious.

So you start the familiar routine: squinting at letter shapes, googling “geometric sans serif lowercase g,” uploading to various font identification sites that give you 20 wildly different guesses, or posting to design forums hoping someone recognizes it.

There had to be a better way.

Introducing What’s That Font

What’s That Font is our answer to the font identification problem. Drop any reference image—screenshot, mockup, inspiration shot—and it identifies every typeface in the image, region by region.

Not just a single best guess. Not a vague category. Specific fonts with:

Confidence scores - How certain the identification is
Visual rationale - Why this font matches what’s in the image
Direct links - Google Fonts embed codes or purchase/download links
Region-by-region analysis - Different fonts in different parts of your image? We catch them all

Try it: whatsthatfont.com
Source: github.com/lab-zee/WhatsThatFont

Powered by Vision AI

Under the hood, What’s That Font uses Claude Opus 4.6 vision—Anthropic’s multimodal model that can analyze images with sophisticated understanding of typography, layout, and visual design.

The workflow:

Upload your image (JPG, PNG, WebP up to 10MB)
Claude analyzes the image region by region
Identifies typefaces based on letterforms, weights, styles
Returns ranked results with confidence scores
Provides visual rationale explaining the matches
Links directly to Google Fonts or foundry pages

All processing happens in-memory. Your images are never stored, logged, or sent anywhere else. Upload, identify, done.

Why We Built This

Lab Z works across multiple domains—AI systems, data tools, design—and we constantly reference visual inspiration. Font identification kept coming up:

“What font is that in their marketing page?”
“Can we match this typeface from our brand guidelines?”
“I saw this in a screenshot—what is it?”

Existing tools were hit-or-miss. Some worked for clean, high-contrast text. Most struggled with:

Multiple fonts in one image
Low contrast or stylized treatments
Similar-looking typefaces requiring nuanced distinction
Explaining why a match was suggested

We had access to Claude’s vision capabilities and realized: this is exactly the kind of visual reasoning task that modern AI excels at.

Open Source, Privacy-First

Like all our tools, What’s That Font is MIT licensed and completely open source.

You can:

✅ Use it freely at whatsthatfont.com
✅ Self-host for your team
✅ Fork and customize
✅ Run it locally during development
✅ Contribute improvements

Privacy guarantee: Images are processed in-memory and never persisted to disk, cloud storage, or logs. We don’t track what you upload. Your visual research stays private.

Tech Stack

For those interested in the architecture:

Framework: Next.js 16 + React 19 + TypeScript
Styling: Tailwind 4
AI: Claude Opus 4.6 via Anthropic SDK
Fonts: Google Fonts API + curated catalog
Rate limiting: Pluggable (memory/Upstash/off)
Deployment: Docker-ready, Railway/Vercel compatible

Full details in the repo’s architecture docs.

Accuracy & Testing

We take accuracy seriously. The repo includes:

Evaluation framework - Generate labeled test images with known fonts
Baseline comparisons - Track accuracy improvements over time
Real-world fixtures - Test cases from actual design work
TDD approach - New features require tests first

The eval suite uses Playwright to render fonts to PNGs with known ground truth, then scores identification accuracy against the Anthropic API.

Get Started

Use Online

Visit whatsthatfont.com and drop an image. That’s it.

Run Locally

git clone https://github.com/lab-zee/WhatsThatFont
cd WhatsThatFont
cp .env.example .env
# Add your ANTHROPIC_API_KEY
pnpm install
pnpm dev

Visit http://localhost:3000

Docker Deployment

docker compose -f docker/docker-compose.yml up --build

Single command, no Node setup required.

Use Cases

Designers: Identify fonts from mood boards, competitor sites, or inspiration screenshots.

Developers: Match typefaces from design mockups when Google Fonts embed codes aren’t provided.

Marketers: Ensure brand consistency by verifying fonts in external materials.

Hobbyists: Satisfy curiosity about typefaces you encounter in the wild.

Contributing

We welcome contributions:

Bug reports - Found an incorrect identification? Let us know
Test cases - Share challenging images that should work better
Font catalog - Suggest additions to our curated list
Code improvements - PRs welcome

See CONTRIBUTING.md for details.

The Bigger Picture

What’s That Font reflects our approach at Lab Z: leverage AI for tasks humans find tedious but machines can automate.

Font identification is pattern matching at scale. Humans can do it, but it requires experience, time, and often trial and error. AI vision models can analyze letterforms, compare against thousands of fonts, and explain their reasoning—all in seconds.

This frees designers to focus on creative decisions rather than detective work.

Not Just for Data Science

You might notice this project isn’t about distributed intelligence, strategic decision-making, or ambient AI systems like our other work.

That’s intentional.

Lab Z builds tools we wish existed—whether they’re infrastructure for high-state intelligence systems or simple utilities that solve everyday friction. Open source means sharing solutions, not just research.

If you’ve ever squinted at a screenshot wondering “what font is that?”, this one’s for you.

Try it: whatsthatfont.com
Source: github.com/lab-zee/WhatsThatFont
License: MIT

Drop an image and see what you discover. We’d love feedback on accuracy, features, or fonts you’d like added to the catalog.