Turn historical documents into research data — with AI you can train yourself.
Train custom AI models on your specific sources, measure accuracy with Character Error Rate, and publish reproducible results. Used at over 3,000 institutions for research, teaching, and library digitization. EU-hosted, GDPR-compliant, and built by a European cooperative with academic roots.
Used at Cambridge, Edinburgh, Helsinki, Radboud, Innsbruck, and 3,000+ higher education institutions worldwide.
What's your starting point?
Whether you're running a research project, teaching a course, or managing a library — here's how Transkribus fits into your work.
Build datasets from primary sources — at a scale that wasn't possible before.
Train a custom AI model on your specific sources, measure its accuracy with Character Error Rate, and document every step for your methodology section. Your models are versioned, your results are reproducible, and your training data is yours. Proven in ERC, FWF, DFG, and AHRC-funded projects.
- Train custom models on your scripts: Secretary hand, Kurrent, Ancient Greek, colonial Spanish, Latin …
- Measurable accuracy: CER as your standard metric for peer review
- Versioned models — share publicly with the community or keep private to your group
- EU-hosted, GDPR-compliant — meets the strictest institutional data requirements
Build datasets from primary sources — at a scale that wasn't possible before.
Train a custom AI model on your specific sources, measure its accuracy with Character Error Rate, and document every step for your methodology section. Your models are versioned, your results are reproducible, and your training data is yours. Proven in ERC, FWF, DFG, and AHRC-funded projects.
- Train custom models on your scripts: Secretary hand, Kurrent, Ancient Greek, colonial Spanish, Latin …
- Measurable accuracy: CER as your standard metric for peer review
- Versioned models — share publicly with the community or keep private to your group
- EU-hosted, GDPR-compliant — meets the strictest institutional data requirements
Radboud University: Transkribus as an institutional service
Deciphering a secret diary in Ancient Greek — with a START Grant
For researchers
Train your own AI — reproducible, publishable, yours.
Research projects powered by Transkribus
From English wills to Spanish colonial fleets — see how university researchers use AI to tackle ambitious digitization projects.
25,000 English Wills Transcribed
University of Exeter and The National Archives — Material Culture of Wills project with 2,500+ Zooniverse volunteers.
Unlocking the New Spain Fleets
Lancaster University — millions of colonial Spanish maritime trade documents from Mexico and Seville archives.
The Egerton Model for Secretary Hand
Northwestern University — 3% error rate on 750,000 words of 16th-17th century English equity court material.
Transcribing Ancient Greek
Cornell, Cambridge, and Innsbruck — three projects pushing AI text recognition into classical and ancient scripts.
Published university research collections
University research projects that publish their transcribed collections as searchable Transkribus Sites — open access, citable, and permanent.
For decision-makers
Bring Transkribus to your entire university
Already using Transkribus? Help your university join.
Many university adoptions start with a single researcher. If Transkribus is already part of your workflow, we'll help you build the case — with usage data, institutional examples, and a contact for your IT or research support team.