Transkribus
Platform from Innsbruck that turns handwriting and historical prints into searchable text with AI, including models you train yourself.
What it does
Transkribus reads scans and phone photos of handwritten or printed documents and turns them into searchable text. It supports more than 100 languages and historical scripts. The platform detects page layouts, tables and fields. In the editor you correct and tag the results, including together with a team. You can train your own recognition models or use one of more than 300 community models. Transkribus Sites lets you publish searchable editions. From the Scholar plan up, an API returns JSON, PAGE XML or ALTO.
Who it suits
Archives, libraries, research projects and genealogists. It also fits companies that want to make old files, parish registers or handwritten records searchable.
Data protection for businesses
Data is stored on the vendor’s own servers in Austria, and the API processes data in Austria and Germany. If you want to upload personal data, you must first sign a data processing agreement (Data Processing Addendum). Under the terms of use, the vendor may also use your content to improve its services. The vendor does not mention a way to switch this off. It shares models and training data only with your consent. Institutions can get the Organisation plan on individual quote.
Limits
Billing is per page via credits, and the allowances in the smaller plans run out quickly. Special categories of personal data, such as health data, are excluded. With hard-to-read originals, you will need to correct the results.