Note Scanned pages, books, forms, tables and formulas. Roughly ordered by recent releases, useful updates and current usage; some models need a full parsing pipeline.

Note Language-specific OCR, including Thai, Japanese, Vietnamese, Arabic, Korean and Devanagari. Check each model's expected input and domain.

Note Handwriting and historical print, including Swedish, Norwegian, German Kurrent, Tibetan and Hebrew-script manuscripts. Many models need cropped lines.

Note Text-line and region recognition, including Kraken and PaddleOCR, plus pipelines that find and read text in documents and photographs.

Note Where models are scored: benchmark datasets with results tables, then per-language leaderboard Spaces.