Transkribus

Overview

Transkribus is a document-recognition workspace for turning handwritten and printed pages into searchable text. It is best suited to researchers, archives, and teams handling historical material who need to refine transcription and page structure, not just receive raw OCR output. The trade-off is a credit-based workflow, and API access starts at the paid Scholar tier.

READ-COOP SCE, a member-owned cooperative founded in Innsbruck in 2019, makes the service. The maker says recognition covers more than 100 languages and historical scripts. Documents are processed and stored on servers in Austria under GDPR-compliant handling, according to the maker; users retain ownership of documents and transcriptions and can delete them.

Key features

Recognition and models: Transkribus handles handwriting and print, with public models for handwriting, print, layout analysis, and tables. Users can also train their own models, which makes it more adaptable to specialized collections than a fixed recognition workflow.

Layout analysis and editing: The service detects text regions, baselines, tables, and reading order. Its editor supports text and layout changes, version history, and real-time collaboration, allowing a project to move from recognition into review without leaving the workspace.

Exports: Output includes PAGE XML, ALTO, searchable PDF, and other formats. This breadth is useful when a project needs both structured layout data and readable documents.

API: The beta Developer Platform API accepts document images and returns structured text as JSON, a ZIP archive, or PAGE XML. Submitted JPEG, PNG, or TIFF files can be up to 20 MB, and results are retained for 24 hours. The Scholar-or-higher gate makes it a poor fit for someone seeking a free API trial.

Credit use: Handwriting recognition, table recognition, and field recognition each cost 1 credit per page; line detection costs 0.25 credits per page. Heavy processing can therefore exhaust a small allowance quickly.

Desktop: The older eXpert application supports Windows, macOS, and Linux, but is no longer updated; new features are available only in the web app. Desktop users should treat it as a legacy option rather than the current product.

Pricing

The free plan costs nothing, billed annually, and includes 50 credits per month, one user seat, 20 GB of document storage, and five training runs per month. It is enough to explore the workflow or process a modest collection, but the monthly credits constrain sustained use.

Scholar costs €99 a year, equivalent to €8.25/month, and includes 900 credits per year, one seat, 200 GB storage, and 30 training runs per month. Prices include 20% Austrian VAT. It is the entry point for API access and suits an individual with regular workloads; the annual credit pool is still limited for large-scale processing.

Team costs €449 a year, equivalent to €37.42/month, including 1,500 credits per year, five seats, 1 TB storage, and 100 training runs per month. It is the better fit for a small collaborative project, though its credit allowance is not dramatically larger than Scholar’s despite the higher price.

Organisation has custom pricing, custom user seats and storage, and API access. For Organisations is also custom pricing, with a quote, 15+ seats, 75,000+ credits per year, a dedicated success manager, and custom onboarding; it is aimed at larger deployments rather than individual researchers.

A one-time Credit top-up costs €5,950, paid once for 250 credits, requires no subscription, and the on-demand credits never expire. It can cover occasional extra processing, but it does not replace the storage or seats of a plan.

Platforms

Transkribus is a cloud service available on the web, with API access and legacy desktop software for Windows, macOS, and Linux. The desktop app’s lack of updates makes the web app the practical choice for current features.

Who it's for

Choose Transkribus for historical documents, handwriting, and projects where recognition needs to be corrected, organized by layout, and exported in structured formats. Its model library and training option suit collections with specialized material; collaboration and higher storage make Team more suitable than Scholar when several people need to work together. If the need is only a low-cost OCR endpoint, the paid API threshold and credit accounting are drawbacks.

Pros and cons

Pro: Recognition, layout editing, version history, collaboration, and structured exports sit in one workflow, reducing the need to assemble separate tools for transcription and page structure.

Pro: Users can choose public models or train their own, which supports work on varied handwriting and historical material.

Pro: The free plan includes 50 monthly credits, and top-up credits do not expire.

Con: Processing consumes credits page by page, with several recognition modes costing a full credit per page.

Con: API access requires Scholar or above, and the API remains in beta.

Con: The legacy desktop application is no longer updated, so new features require the web app.

Alternatives

For a broader set of OCR options, compare the AI OCR Software directory; for document-focused workflows, see Document Management Software.

Doxis AI.dp is a paid alternative with free credits and pay-as-you-go billing, suitable for readers who prefer usage billing over Transkribus’s credit plans. Veryfi API offers a freemium API plan of up to 100 documents per month, making it worth considering for a bounded API workload. Nanonets Accounts Payable Automation is another freemium web-and-API option. Baidu OCR is a paid OCR option with Android, iOS, Linux, web, and Windows platforms and a one-time 10万点 package with a 12-month validity period. Cloudmersive OCR API has a free tier of 600 calls per month, limited to 1 call per second and 3.5 MB files, for readers who need a small API allowance. Instabase offers a community trial with $5 of usage that ends after the included units are used or 92 days. Amazon Textract AnalyzeExpense has a new-customer free tier of up to 100 pages per month for three months. Reducto starts with $150 in free usage and then pay-as-you-go billing.

Verdict

Transkribus is the strongest fit for people working with historical handwriting who need recognition, layout correction, model choice, and export in one cloud workspace. Its free tier gives an individual a way to begin, while Scholar unlocks API access; look elsewhere if you need an unrestricted free API or predictable high-volume processing without page-credit limits.