Donut

Donut (Document Understanding Transformer) is an OCR-free, end-to-end transformer for visual document understanding from Naver Clova. Rather than chaining an external OCR engine to a language model, Donut reads the document image directly and generates structured output, reaching state-of-the-art results on visual document classification and information extraction tasks like receipt parsing.

The work also introduced SynthDoG, a synthetic document generator that makes pretraining flexible across languages and domains. The paper appeared at ECCV 2022 (arXiv 2111.15664). Donut became a foundational reference for the OCR-free document AI line that later models like Nougat built on; verify the repository license before commercial deployment.