From illegible handwriting
to searchable text.
BIQE AI HTR SaaS corrects raw transcriptions of historical manuscripts into fluent, contextually accurate text — ready for cataloguing, indexing or publication. Cloud-delivered, pay-as-you-go, no upfront investment.
Also available in Dutch, German and French — contact us.
Who is this for?
Archives
Municipal, provincial and regional archives with registers, minutes and administrative documents from the 17th century onwards.
Museums and heritage institutions
Collections of personal correspondence, diaries and historical documents that need to be made searchable.
Universities
Research projects that need to systematically transcribe large volumes of manuscripts, without investing in their own infrastructure.
Scanning and service bureaus
Digitisation firms that want to offer their clients a complete workflow — from scan to searchable text — without building in-house LLM expertise.
Hospitals and healthcare institutions
Digitisation of historical medical records and patient archives, with strict requirements for delivery and data processing.
Legal firms
Law firms and notaries with case files from before the digital era, now needing to be better accessible for case research and heritage documentation.
How it works
BIQE AI HTR SaaS is a correction layer, not a complete new pipeline. You run layout analysis and character recognition on your own hardware — software you likely already know. Our service takes over at the point where most models fail: semantic correction based on contextual understanding.
Layout analysis and HTR on your own GPU. Scans stay with you.
One image + raw transcription per page to our API. Encrypted via HTTPS.
Corrected text returned within seconds. Coordinates unchanged — ready for ALTO/PDF export.
What makes this different?
Tuned for historical material
Our prompts are specifically developed for 17th–20th century handwriting, primarily Dutch but also English, German and French. For 19th-century administrative documents we achieve error rates of 2–3% at the line level.
Model-agnostic
Behind the scenes we automatically choose between Gemini, Claude and other LLMs based on the quality tier you select. You don't need to worry about model versions or providers.
No data retention
Scans are processed per page and not stored. Only anonymous statistics (page count, compute time) are kept for billing and quality monitoring.
Predictable costs
Three clear tiers, transparent per-page pricing, no surprises. Monthly limits to prevent budget overruns.
Flexible integration
Works as a plug-in with any HTR pipeline that produces PageXML. Your existing processes and output formats (PageXML, ALTO, searchable PDF) remain unchanged.
Multilingual support
Developed and delivered from the Netherlands, with direct support in Dutch and English. Available for clients across Europe and the United States.
Interested?
We like to start with a free pilot of 100 pages on your own material. Then we can see together whether the quality fits your collection.