Researchers at Google have developed a new method for converting photos of handwriting into digital ink, allowing for a return to handwritten notes. The Large-i model, part of the team's "Large-i" framework, can learn to read and write handwritten text without paired training data. This approach is robust to various inputs, including full handwritten notes and out-of-domain sketches. Human evaluation results show that the Large-i model performs similarly to human-generated digital inks, with 87% of its outputs considered good or having only small errors.

Source: https://research.google/blog/a-return-to-hand-written-notes-by-learning-to-read-write/