GlyphForge

From records
to trees.

Machine learning that reads any historical record — handwritten documents, journals, parish books, civil registers — extracts names, dates, and relationships, and builds genealogical trees automatically.

HTR — handwriting transcriptionOCR — page extractionGraph — lineage explorerExport — GEDCOM
Orion →See how it works

The graph, moving.

Explore a living family tree

Drag, hover, and zoom through a sample family spanning five generations. This is a taste of what GlyphForge builds automatically.

Loading graph…
drag Drag to rearrangehover Hover for detailszoom Scroll to zoom

What's inside.

  • 01
    Relationship Extraction
    Detects parent-child, spouse, godparent, and sibling relationships from contextual clues in historical records.
  • 02
    Handwriting Transcription
    Purpose-built models that read centuries-old cursive, fraktur, and degraded manuscripts with high accuracy.
  • 03
    OCR & Document Intelligence
    Robust optical character recognition for printed and handwritten text across parish registers, civil records, and censuses.
  • 04
    Data Extraction
    Automatically identifies and structures names, dates, places, and life events from unstructured document text.
  • 05
    Records to Trees
    Connects extracted individuals across thousands of pages into validated, deduplicated family trees ready for research.

How it works.

  1. 01
    Upload

    Scanned parish records, civil registers, or family books in PDF format.

  2. 02
    Extract

    HTR and NLP models read handwriting and identify people, dates, and relationships.

  3. 03
    Build Graph

    Extracted data is ingested into Neo4j and validated with 30+ integrity checks.

  4. 04
    Explore

    Visualize the family tree in an interactive force-directed graph at 60fps.