Transkribus

Overview

Transkribus is a cloud service that turns images of handwritten and printed documents into searchable text. It is best suited to researchers, archives, and organisations working with historical material who need recognition, layout editing, and document management in one workflow. Its free tier is useful for small projects, but processing limits and the paid API threshold matter as work grows.

Made by member-owned European cooperative READ-COOP SCE, Transkribus combines AI recognition with public and trainable models, editing tools, and structured exports. The maker says it supports more than 100 languages and historical scripts. The service processes and stores data on servers in Austria with GDPR-compliant handling; users keep ownership of their documents and transcriptions and can delete them at any time.

Key features

• Handwriting and print recognition: Recognition across more than 100 languages and historical scripts makes the service relevant to mixed and older collections, rather than only clean, modern print. Results still consume credits, so high-volume projects need to budget for usage.
• Public and custom models: Users can browse public models for handwriting, print, layout analysis, and table recognition, or train their own. That flexibility suits collections with specialised material, while the monthly training-run allowance varies by plan.
• Layout analysis and export: The system detects text regions, baselines, tables, and reading order, then exports to PAGE XML, ALTO, searchable PDF, and other formats. This is useful when preserving page structure matters as much as extracting words.
• Editing and collaboration: Text and layout editing, version history, and real-time collaboration keep correction and review alongside recognition. The web app is where new features are available; the older eXpert desktop software for Windows, macOS, and Linux is no longer updated.
• Developer API: The beta API accepts document images and returns structured text as JSON, a ZIP archive, or PAGE XML. It requires Scholar or above, retains results for 24 hours, and accepts JPEG, PNG, or TIFF images up to 20 MB. Those constraints suit short-lived integration workflows better than a permanent result store.
• Credit-based processing: Handwriting recognition, table recognition, and field recognition each use 1 credit per page; line detection uses 0.25 credits per page. The differing rates make the included allowance dependent on the work being run, not just the number of pages.

Pricing

Transkribus is freemium. Prices for Scholar and Team include 20% Austrian VAT, and annual allowances should be weighed against page-based credit consumption.

• Free: free, billed annually, with 50 free credits per month, 1 user seat, 20 GB document storage, and 5 training runs per month. It is a sensible starting point for individual experiments, but the monthly credit cap and single seat constrain sustained projects.
• Scholar: €99 a year, equivalent to €8.25/month, with 900 credits per year, 1 user seat, 200 GB storage, and 30 training runs per month. This is the entry point for API access and a practical upgrade for an individual with regular processing needs.
• Team: €449 a year, equivalent to €37.42/month, with 1,500 credits per year, 5 user seats, 1 TB storage, and 100 training runs per month. It fits collaborative projects that need shared capacity, though its credit pool is not unlimited.
• Organisation: Custom pricing, custom user seats and storage, and API access. Larger organisations can request a configuration rather than work within the fixed seat and storage limits.
• For Organisations: Custom pricing for 15+ user seats and 75,000+ credits per year, with a dedicated success manager and custom onboarding. This is aimed at larger deployments rather than small research groups.
• Credit top-up: €5,950, paid once for 250 credits; no subscription is required, and on-demand credits never expire. This offers a way to extend usage without changing plans, but the listed purchase is a substantial one-off commitment.

Free includes standard support; Scholar and Team include priority support. The fixed plans keep seat, storage, and processing allowances clear, but API users cannot stay on Free and heavier workloads must account for credit use.

Platforms

Transkribus is available on the web and through an API, with Linux, macOS, and Windows support for the older eXpert desktop software. Because that desktop application no longer receives updates, users who want current features should work in the web app. Deployment is cloud-based.

Who it's for

Transkribus is a strong fit for archives, researchers, and teams digitising historical documents, especially when they need handwriting recognition, page-structure analysis, correction tools, and exports together. It is less suitable for an API-first user who needs to start without a paid plan, or for a project whose processing volume does not fit the available credit allowances.

Pros and cons

Pros

• Recognition supports more than 100 languages and historical scripts, according to the maker, broadening its usefulness for archival material.
• Model browsing and training, layout analysis, editing, version history, and collaboration cover more of a document project than recognition alone.
• Users retain ownership of documents and transcriptions and can delete them; the maker says processing and storage are in Austria with GDPR-compliant handling.
• The free plan includes 20 GB storage and 5 training runs monthly, giving individuals room to assess a project before paying.

Cons

• Credits are consumed per page, with different rates by processing type, so the free monthly allocation may not go far for large collections.
• The API is beta and locked to Scholar or higher, limiting no-cost integration work.
• The older desktop application receives no updates, making the web app the practical choice for new features.
• The 24-hour API result retention and 20 MB image limit may not suit workflows that need longer-lived results or larger source images.

Alternatives

Consider Veryfi API instead when an API-focused document workflow with a stated free allowance is a better fit: its Free plan covers up to 100 documents per month across document types, with limited storage and email support.

Base64.ai may suit users looking for OCR in 165+ languages, handwriting support, API access, and unlimited free seats; its listed 1 cent OCR plan is $0.01 a month.

For a stated no-expiration API free tier, Cloudmersive OCR API offers 600 calls per month, a limit of 1 call per second, and a 3.5 MB maximum file size.

Mistral OCR is another API-oriented option, with OCR 4.1 API priced at $4 a month, billed at $4 / 1000 pages.

For enterprise accounts payable automation, consider Hyland AP Automation.

LandingAI Agentic Document Extraction may suit teams seeking an Explore tier with 1 user seat and 1,000 free credits to start, plus a stated Team plan.

Nanonets Accounts Payable Automation is another option for accounts payable automation, with a Starter plan for up to 3 users.

Doxis AI.dp offers €25.00 worth of free credits on a pay-as-you-go basis, with usage billed monthly.

Browse more options in AI OCR Software or Document Management Software.

Verdict

Choose Transkribus for historical-document projects that need handwriting recognition alongside layout analysis, editing, collaboration, and structured export. Its breadth and free entry tier make it a credible starting point, but credit caps and the Scholar minimum for API access are reasons to look elsewhere if the priority is high-volume or no-cost API processing.