danilo-znamerovszkij/draw-your-font

What it solves

It converts photographs of handwritten alphabets into installable font files (TTF, WOFF, WOFF2), eliminating the need for paid services or manual digitizing. It handles messy real-world photos—including those with bad lighting or spiral notebook bindings—and processes everything locally for privacy.

How it works

The tool uses a pipeline that converts a photo into a font through several steps:

1. Processing: It uses adaptive thresholding and blob detection to isolate handwritten characters.
2. Labeling: Characters are identified either via a vision-capable AI (like Claude), a predefined template order, or manual user input.
3. Vectorization: It uses the potrace engine to turn images into vectors.
4. Metrics: It maps characters to a shared 1000-unit em square to ensure consistent cap height, x-height, and descender depth, making the result feel like a professional font.

Who it’s for

People who want to digitize their own handwriting for personal or commercial use without paying for subscriptions or uploading their data to external servers.

Highlights

• AI-Integrated: Can be used as a Claude Code skill to automatically label letters and iteratively refine the font through conversation.
• Local Processing: All processing happens on the user's machine; no data is uploaded.
• Zero Dependencies: Runs on Node.js without requiring external binaries like FontForge or ImageMagick.
• Flexible Input: Supports both structured printable templates and freeform photos (e.g., napkins or notebooks).