Skip to content

Roadmap ​

Everything below works in the Transkribus app today; the middle column links to the help center article. Text recognition, sites, datasets, projects and credits have their own APIs on the platform; stored material is the Transkribus REST API, documented under Collections, Documents, Analyse documents, Models and Search. Each new platform surface is announced in the changelog when it ships.

CapabilityIn the appAPI status
Recognise text: turns an image of handwritten or printed text into a transcription with coordinates for every region, line and word.Transcribing documentsAvailable, see Recognise text.
Analyse layout: detects the structure of a page, text regions, baselines, line segmentation and reading order. Layout models produce this geometry independently of transcription.Layout recognitionInside text recognition jobs via config.lineDetection; on stored documents see Analyse documents. No standalone single-image job yet.
Recognise tables: detects table structure on a page and returns cells with their row and column positions, so tabular material comes back as data rather than a flat sequence of lines.TablesNot available through the API yet.
Extract fields: detects regions of a page against a template you define once and returns those regions filled for every page of the same type. Suited to registers, forms and record series where the layout repeats.Field modelsNot available through the API yet.
Train models: trains a model from ground-truth material and registers the result so it can be used by id like any other model.Training modelsAvailable, see Models.
Browse models: lists what is available with language, script, material type and quality, and whether a model is public or belongs to your account.Managing modelsAvailable, see Models; public models on the Transkribus website.
Collections: a set of documents that share models, permissions and settings; the unit most institutional workflows are organised around.Managing collectionsAvailable, see Collections.
Documents and pages: an ordered sequence of pages, each backed by one image and carrying its transcription, layout and any extracted data.Uploading and downloading documentsAvailable, see Documents. Text recognition jobs take a single image in the request instead; see Images.
Projects: bundles collections, models, datasets and people for one research context, so several bodies of work stay separated.Managing projectsAvailable, see Projects.
Export: produces a downloadable artifact from recognised documents. Format is a parameter, not a separate job type: PAGE XML, ALTO, plain text, DOCX and others are the same export with a different value.Uploading and downloading documentsAvailable for stored documents, see Documents; per job: ZIP, PAGE XML and ALTO XML, see Recognise text.
Datasets: a curated set of transcribed pages, used as training data or published as a citable research output. Datasets close the loop: corrected results become training material, which becomes models that perform better on your documents.Help centerAvailable, see Datasets.
Sites: a public, searchable presentation of a collection, so third parties can build on what an institution has made public.Transkribus SitesAvailable, see Sites.

Transkribus Developer Platform