# Show HN: Photo of a handwritten alphabet to installable TTF, open source

> Source: <https://github.com/danilo-znamerovszkij/draw-your-font>
> Published: 2026-07-24 08:28:37+00:00

**Turn a photo of your handwriting into a real font (TTF/WOFF/WOFF2) - free, open source, no uploads, no credits.**

Draw your alphabet on paper. Take a photo. Get your font.

*This is a real one-shot result: dim light, spiral binding, page shadow. One photo in, installable font out.*

```
npx skills add danilo-znamerovszkij/draw-your-font
```

Then in Claude Code, just talk:

"here's a photo of my handwriting - make my font"(drag the photo into the terminal)

Claude finds your letters in the photo, labels them with vision, builds the font, shows you a preview, and critiques its own work. Iterate by talking:

*"make it rounder"*/*"a bit bolder"**"the g looks bad"*- it shows you the crop and fixes or asks for a re-shoot*"give me woff2 + css for my website"**"how readable is it?"*- a legibility score and the two worst letter pairs

No photo yet? Say *"give me a font template"* and you get a printable PDF grid:
write your alphabet with a dark pen, photograph the pages, and hand them back.
Messy freeform photos work too. Napkins, notebooks, spiral binding, bad
lighting: that's what the vision step is for.

Everything runs locally on your machine. Your handwriting never leaves it.

The skill is a thin layer over a deterministic npm CLI. It works on its own when you can tell it what you wrote:

```
# freeform photo, you know the order you wrote in:
npx draw-your-font make photo.jpg --chars "ABCabc" --name "My Hand"
# → MyHand.ttf - double-click, install, done.

# best quality: print a template, fill it, photograph:
npx draw-your-font template -o template.pdf --charset minimal   # or: spanish
npx draw-your-font make page1.jpg page2.jpg --charset minimal --name "My Hand"
```

Pure npm, zero system dependencies: no FontForge, no ImageMagick, no potrace binary. Works on macOS / Linux / Windows wherever Node ≥ 18 runs.

| Command | What it does |
|---|---|
`template` |
printable A4 PDF grid (`--charset minimal|spanish` ) |
`segment <photos…>` |
find letters → crops + numbered contact sheet + `blobs.json` |
`build` |
labeled crops → font (`--labels` / `--chars` / `--charset` ) |
`make <photos…>` |
segment + build in one shot |
`preview` |
render any text with the built font |

Refinement flags: `--smooth 0..2`

(rounder curves), `--weight=-2..2`

(thinner/bolder), `--formats ttf,woff,woff2,css`

(web-ready + `@font-face`

snippet). Run `draw-your-font --help`

for everything.

```
photo ──► adaptive threshold ──► blob detection ──► label (Claude / you / template order)
      ──► potrace vectorize ──► shared em-square metrics ──► TTF/WOFF/WOFF2 + preview
```

The craft is in the metrics step: every character has a vertical band in a
shared 1000-unit em square (cap height, x-height, descender depth), so your
`g`

hangs below the line and your `o`

stays small. That's what makes it feel
like a font instead of a ransom note. Vectorization is potrace, the same
engine inside FontForge and Inkscape. AI never draws your letters; it only
finds, labels, and judges them.

**Who owns the font?** You. 100%, commercial use included. It's your
handwriting.

**Why is this free when Calligraphr charges $8/month?** Their cost is
servers and a browser editor. Here your machine does the work and the agent
is the editor.

**Kerning, ligatures, letter randomization?** v2. The pipeline (fonttools
`calt`

) is planned; the current output is a clean single-variant font.

MIT. Draw something.