One-DM

Why we included this project

One-DM generates handwritten text in a specific person's style from a single sample of their writing. The diffusion model extracts the style from that one reference and renders whatever content you give it, and the authors report it beats earlier methods that needed many more samples. It was tested on English, Chinese, and Japanese handwriting. The repo is a PyTorch implementation tied to an ECCV 2024 paper, with the paper, poster, and demo video linked in the README. It's research code rather than a polished application, so expect to work with the training and inference scripts directly. Researchers working on handwriting synthesis or synthetic data for OCR training are the main audience.