mittagessen/kraken

Health v2 · maintenance only

Adoption not part of the score

1061 stars · 173 forks observed · 2026-09-01

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

kraken is a turn-key OCR/HTR engine built on neural networks, optimized for historical and non-Latin script material. It provides trainable layout analysis, reading order detection, and multi-script character recognition with output in ALTO, PageXML, abbyyXML, and hOCR formats.

Use cases

• recognize text in historical manuscripts
• OCR for non-Latin scripts like Arabic or Hebrew
• transcribe handwritten documents
• extract text from scanned images into PageXML or ALTO
• binarize and segment scanned page images
• train custom OCR models for rare scripts
• handle right-to-left and bidirectional text in OCR

When to choose

• you need OCR for historical, damaged, or non-Latin script material
• you want fully trainable recognition and layout analysis models
• you need structured output formats like ALTO, PageXML, or hOCR
• you work with right-to-left, BiDi, or vertical scripts
• you need both printed text and handwritten text recognition

When to avoid

• you need a plug-and-play OCR for modern Latin-script documents with no training
• you want a GUI-based OCR tool
• you need Windows support
• you want commercial support or prebuilt binaries outside pip

Facets

library · maturity active

ocr machine-learning image-processing nlp computer-vision pdf files python cli handwritten-text-recognition htr layout-analysis pagexml alto-xml hocr historical-documents non-latin-scripts neural-networks natural-language-processing linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="mittagessen/kraken")