Handwriting recognition has come a long way, but it still struggles with the writing that matters most. People with dysgraphia and other handwriting difficulties often find that even the best OCR systems can't reliably read their work. This isn't surprising—these systems are trained on neat, consistent handwriting, not the varied and unconventional styles that real people actually write in.
DisgraPhi takes a different approach. Instead of trying to build one model that works for everyone, it learns your handwriting specifically. You provide a small sample of your writing, and the system fine-tunes itself to understand your unique style. The result is a personalized OCR model that actually works with how you write, not against it.
The system builds on modern vision-language models but keeps things practical. It's designed to run on consumer hardware and doesn't require massive datasets or expensive infrastructure. A few pages of your handwriting are enough to train a model that understands your writing better than any generic solution could.
This matters because between 5-20% of people struggle with handwriting in some way, and that number keeps growing. Whether it's dysgraphia, motor difficulties, or just unconventional penmanship, everyone deserves tools that work with their writing. DisgraPhi is a step toward making that happen.