Transkribus Platform

11/06/2026

Can a handwriting recognition platform read text that was never written in ink?

Most AI-powered text recognition systems are built to interpret visual marks on a page. But Braille was never designed to be seen, it was designed to be felt.

At the TUC 2026, Ellen Forget from the University of Alberta will share how she trained a Braille character recognition model in Transkribus, tackling the unique challenge of teaching AI to recognise tactile text.

The project highlights how established HTR technologies can be extended beyond traditional handwritten texts, creating new opportunities to digitise Braille collections and advancing the development of an open-access recognition model for Braille documents.

👉 Join us at to discover how expanding the definition of "text" can expand access to knowledge:

Panel 1Lightning TalksTuesday 22 September10:00 – 11:15 Presented by Ellen ForgetUniversity of AlbertaTraining Braille Character Recognition Using TranskribusWhat happens when you ask a handwriting recognition platform to read text with no ink at all? This talk charts the development of a Braille ...

08/06/2026

Many archives now work with AI-generated transcriptions. Far fewer have a workflow for improving them.

Many national and state archives already have engaged communities willing to help transcribe handwritten collections. But once the text is published, there’s often no easy way for users to correct mistakes or contribute improvements back into the system.

The Swedish National Archives (Riksarkivet) is testing a different approach.

Through a pilot integration between its national search service (NAD) and Transkribus, users can open archival files, correct transcriptions, and sync those improvements directly back into the public portal using IIIF.

Instead of treating transcription as a finished product, the workflow turns archival users into active collaborators while also generating valuable ground truth data for future HTR models. Meaning, the project connects human expertise and automated text recognition in a workflow designed to scale.

If you’d like to learn how a major national archive is approaching collaborative HTR workflows in practice, join the lightning talk by Linnéa Karlberg Lundin from the Swedish National Archives at the Transkribus User Conference 2026 this September.

👉 View the programme and secure your ticket:

Panel 6Lightning TalksTuesday 22 September13:45 – 15:00 Presented by Linnéa Karlberg LundinThe Swedish National ArchivesUsing Transkribus to bring community contributions into the HTR workflow at the Swedish National ArchivesMillions of archival documents, transcribed by HTR — but with no way f...

03/06/2026

It's not every day a digital humanities project receives an award presented by Queen Mary of Denmark!

Congratulations to Professor Johan Heinsen and his team at Aalborg University - their project, Enevældens Nyheder Online, was recently awarded the prestigious "Original Idea of the Year" award by the Independent Research Fund Denmark.

Johan Heinsen’s research tackles a fascinating question: How did the modern state tighten its control over citizens between 1750 and 1850? To find out, his team turned to an enormous archive of historical newspapers to track over 15,000 runaway notices and early crime reports.

Manually searching through half a million pages of old news was impossible for a small team. To make matters worse, traditional methods struggled reading the historical documents, resulting in an error rate too high to be useful for keyword searches.

Using Transkribus, the team trained custom models to handle complex newspaper layouts and recognise historical Danish. This brought the error rate down below 1%, making the entire collection searchable for the first time.
The result is a completely open database that allows researchers and the public to explore hundreds of thousands of pages of history..

We are incredibly proud to see Transkribus used as the foundation for this award-winning work, and we want to thank Johan and his team for sharing their journey with our community!

👉 Read the full award-winning story on our blog:
https://eu1.hubs.ly/H0vSMcJ0

👉 Visit the Enevældens Nyheder Online project website:
https://hislab.quarto.pub/

Explore how Professor Johan Heinsen's award-winning research on social control in Denmark transformed digital humanities through historical newspaper digitisation.

28/05/2026

Transcription used to be the end goal. Now, it’s just the starting point.

Trying to use broad, everyday tech tools for specialised historical work often leads to a major bottleneck. Generic models frequently struggle with complex multi-column layouts, historical handwriting, or unpredictable hallucinations in the text.

At TUC 2026, we’re dedicating three days to discussing how Transkribus and various AI approaches can effectively meet the needs of scholars, archivists, and research institutions in unlocking the written past.

Our sessions will address these practical challenges head on, from building databases with Smart Extract Models and mastering complex layout segmentation to analysing what Visual Language Models can achieve for historical documents.

If you’re working with historical sources and want to better understand what current AI methods can (and cannot) realistically do in practice, join us this September at the Universität Passau!

⏰ Early Bird tickets (25% off) end this Sunday, May 31st.
👉 https://eu1.hubs.ly/H0vHWfj0

27/05/2026

What can Visual Language Models (not) do for document analysis?

There’s no shortage of new AI tools promising to instantly read text, layouts, and visual structures all at once. But when you apply them to complex historical documents, how do they actually perform?

At the Transkribus User Conference 2026, we are going to focus on what actually holds up in practice.

"What Visual Language Models Can (Not) Do for Document Analysis" serves as both the title and the core of Vincent Christlein’s keynote speech, which will kick off our Tuesday sessions at this year's Transkribus User Conference!

As head of the Computer Vision group at the Friedrich-Alexander University of Erlangen-Nürnberg, Vincent Christlein will explore where lightweight and open models excel and where they fall short. He will discuss recent work on structured OCR, information extraction, and document translation to explain why massive end-to-end AI systems might not be the answer, but hybrid pipelines are.

🗓️ Date: 22 September 2026
📍 Location: University of Passau
🔑 Keynote: https://eu1.hubs.ly/H0vG54p0

👉 Take a look at the full programme and join us to explore these questions in more detail: https://eu1.hubs.ly/H0vG5hB0

21/05/2026

Many researchers and archivists run into the same problem: ambitious project goals, strict deadlines, and simply not enough hands to do the work.

The team at the Hargrett Rare Book and Manuscript Library faced exactly this during the Finding Their Names project, which aims to identify and document enslaved individuals mentioned across irregular, handwritten Colonial and Antebellum-era records. The team had to work with scattered documents, strict digital standards, and a very limited timeline. Doing all of this manual processing alone simply wasn’t realistic.

Instead of tackling the massive collection page-by-page, they combined Transkribus with custom scripting workflows. This allowed just two people to generate searchable transcripts and structured text for more than 20,000 pages in under two months.

And, more importantly, the project established a smart workflow that the team can continue using for any future transcription project.

Find out how the right digital tools can help small teams bypass the manual work and make large historical collections far more accessible for research. 👇

Discover how the University of Georgia's innovative project transcribed 20,000 pages of historical documents in just two months, enhancing access to enslaved individuals' stories.

19/05/2026

Digital Humanities is evolving faster than ever. But is it leaving non-Latin scripts behind?

To answer this question, we are excited to welcome Alíz Horváth to the stage at TUC 2026 this September at the University of Passau!

Many researchers working with non-Latin scripts face a frustrating "pragmatic" problem: the technology they rely on isn't always built for the languages they study. This under-representation remains a major barrier to equal access when it comes to AI.

In her keynote, “Why does language inclusivity matter in text recognition and otherwise in the age of AI?”, Alíz Horváth will challenge the current gap in non-Latin script support and explore how we can create an infrastructure that supports every language, not just a select few.

📍 Transkribus User Conference in Passau, Germany
🗓️ 21-23 Sept 2026

👉 Take a look at the full programme and join us to explore these questions in more detail: https://eu1.hubs.ly/H0vqZnd0

13/05/2026

Is your research time being swallowed by tech troubleshooting?

For most historians, "going digital" isn't as simple as it sounds. It usually means a solo mission of hunting for software, managing personal subscriptions, and trying to solve technical hurdles alone. Every hour spent playing IT support is an hour lost to actual research.

Radboud University decided that shouldn't be the norm and moved Transkribus into the library as a central service.
By setting up an Organisation Plan, the university library now handles the technical side and the costs.

This means researchers can stop worrying about the setup and get back to their sources. With the library team there to help, it’s much easier to:
• Find the right model for specific, messy handwriting.
• Make large collections searchable in seconds.
• Get professional tools into the hands of everyone, from students to senior faculty.

Radboud is proving that digital tools work best when they are treated as a shared resource, not an individual burden.
👇 Read the full story of how Radboud is supporting its community:

Discover how Radboud University successfully implemented Transkribus as a comprehensive, university-wide service to bolster digital humanities research.

12/05/2026

History isn’t always about grand narratives; sometimes it's about contraband rhubarb and 17th-century curiosities.

While AI is evolving fast, many general tools still struggle with anything outside of standard English, such as historical German scripts like Kurrent, Sütterlin, and Fraktur. Working with these sources requires models that have been specifically trained for archival material.

Here are three projects where researchers are using Transkribus to get historical German-language collections online:

1️⃣ Scale: The Museum für Naturkunde Berlin transcribed 250,000 handwritten specimen labels, turning millions of data points into a searchable global resource.
2️⃣ Community: The German Archives for Diaries uses a citizen science approach, allowing volunteers to transcribe 27,000 personal diaries and preserve stories from past generations.
3️⃣ Reach: At Latvijas Nacionālais arhīvs, researcher Mairita Lukianska digitised 263,000 pages of Riga’s City Council minutes, a reminder that German-language heritage isn't limited to Germany, and it’s full of surprises (like 17th-century rhubarb crackdowns).

From illegally imported rhubarb to millions of museum specimens, these projects show that with Transkribus, historical German script is no longer a barrier.

👉 Read the full breakdown of these three projects on our blog:

Discover how Transkribus is revolutionising access to historical German-language documents, transforming archives into searchable digital resources for all.

07/05/2026

When is an LLM enough for historical documents — and when isn't it?

That's the core question archivists, DH researchers, and the team behind Transkribus will be answering in Passau from 21–23 September. As AI becomes a standard tool, the real debate isn't about whether to use it, but which approach delivers the precision that messy, centuries-old manuscripts actually require.

👉 View the full programme and plan your participation:
https://eu1.hubs.ly/H0v5CCS0

Here is a look at just two of the specific topics we’re taking on at :

• Building Databases: Get hands-on with Smart Extract Models (SEM). We will look at how to design schemas and turn handwritten sources into structured, database-ready data, and discuss when SEM is (and isn’t) the right approach.
• Abandon All Consistency: Join our roundtable where our speakers will discuss navigating the "methodological hell" of transcription standards in the age of AI.

We’d love for you to be part of this exchange and join us at the University of Passau!

🎟️ Early Bird tickets are still available: https://eu1.hubs.ly/H0v5tlR0