Personal project · local AI

A photo in.
A pattern out.

Crochet turns an amigurumi picture — or a text description — into a beginner pattern with US terms, stitch counts, and a PDF. It runs on your machine. No API bill.


How it works

01

Look

Local vision (Ollama + llava) reads the photo and names the parts: body, ears, wings, leaves.

02

Steer

You review every part — edit shape, color, count — before anything is written.

03

Write

The engine lays out rounds, materials, and assembly, then saves a printable PDF.


The desk

Two columns. Photo on the left. Pattern on the right.

This page is the product face. The live generator runs locally after install — it needs Ollama, so it is not hosted here.


Install

Anyone with the repo can run it

Python 3.10+, Ollama, and the llava model. Full steps live in the README — this is the short path.

git clone https://github.com/GumyGam/crochet-ai.git
cd crochet-ai
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
ollama pull llava
python3 app.py

Then open http://localhost:5001. CLI: python3 image_to_pattern.py photo.jpg "Name"