InkVoice
An accessibility tool that renders generated text in your own handwriting, learned from a sample of your letterforms.
What it is
For people who can't physically write at length but still need handwritten output — a class assignment, a personal note — InkVoice closes the loop. You upload a sample of your handwriting, it extracts your glyphs and spacing statistics, you talk to an LLM about what you want to say, and it renders the result as a PDF in your own hand, with natural variation so it doesn't look stamped.
How it works
A FastAPI backend handles the vision work — glyph extraction and spacing analysis via OpenCV and Pillow — and stores letterforms in SQLite. A Next.js front end runs the sample-upload, chat, and render phases. The hard part is the rendering: sampling among glyph variants and jittering baseline and spacing so a page reads as handwritten rather than as a font.