Search...
  • Pricing
Get Started

Start typing to search across models, sites, and blog posts...

Read Old Handwriting — Free AI Translator

Upload a photo of any old handwritten document. The AI reads it and gives you text you can copy — any script, any century.

Any old script100+ languagesFree to tryNo signup needed

Drag an image here

Select a file...

PNG or JPG up to 10 Mb

Wolpi
AI Assistant

By uploading an image, you accept our terms and privacy policy.

This demo uses Text Titan, our most powerful AI model. Create a free account to get 50 credits every month — no credit card required.

500 000+People use Transkribus
200+ MillionPages of old handwriting read
300+AI models for old scripts
Free50 credits every month

How to read old handwriting with AI

Three steps — no software to install, no account required for the demo.

1

Upload a photo or scan

Take a photo of the old document with your phone or use a scan you already have. Drag the image into the box above. JPG, PNG, and PDF all work.

2

AI reads the handwriting

The AI detects every line of text and converts the old handwriting into digital text — even scripts that most people today cannot read, like Kurrent, secretary hand, or Gothic cursive.

3

Copy, edit, or export

Your text appears in seconds. Copy it, paste it into a document, or create a free account to export as TXT, DOCX, or searchable PDF.

Every century, every script

Old handwriting comes in hundreds of styles. The AI reads them all.

Before the 20th century, people across Europe and the Americas wrote in scripts that look nothing like modern handwriting. Each country, each century, each profession had its own conventions. If you have found an old document — a letter, a record, a page from a book — and cannot read it, the handwriting is probably one of these:
Kurrent & Sütterlin — the angular German scripts used from the 1500s to 1941
Secretary hand — the script of English letters, wills, and records from the 1500s and 1600s
Fraktur & blackletter — the "broken" Gothic print used across Northern Europe
Procesal & Cortesana — dense Spanish notarial hands from the colonial era
Ronde & bâtarde — French formal and legal scripts from the 1400s to the 1700s
Copperplate & round hand — the elegant, looped handwriting of the 1700s and 1800s
Read about Kurrent & Sütterlin
Old handwriting scripts from different centuries and countries

What people decipher

Letters, wills, church records, old diaries — whatever you have

Most people come to Transkribus with one document and one question: what does it say? You might have inherited a box of your grandparents' letters, found a will in the family archive, or stumbled on a parish register entry you need for genealogy research. The AI reads whatever you upload — no need to know which script it is or what century it comes from.
Family letters and postcards in old cursive or Kurrent
Parish registers — baptisms, marriages, and burials in Latin, German, French, or local vernacular
Wills, deeds, and notarial acts in secretary hand or legal scripts
Census records and tax rolls with columns of handwritten entries
Diaries, recipe books, household accounts, and personal journals
Military records, immigration papers, and pension files
Guides for every record type
Old family letters and documents in historical handwriting

Why this works

Purpose-built for historical handwriting — not a chatbot reading images

General-purpose AI tools (ChatGPT, Google Lens, etc.) occasionally guess a word or two from a photo of old handwriting, but they fail on anything beyond simple, modern print. They were never trained on historical scripts. Transkribus is different: it uses specialised neural networks — called Handwritten Text Recognition (HTR) — trained on over 30 million words from real archival documents spanning five centuries. The result is accurate, line-by-line transcription, not guesswork.
HTR models trained on millions of real historical documents, not web text
Layout analysis finds text lines, columns, tables, and marginalia automatically
Handles faded ink, bleed-through, damaged paper, and low-contrast scans
Language models understand old spelling, abbreviations, and vocabulary
300+ public models for specific scripts, languages, and time periods
How the AI works

For larger collections

A whole box of old documents? Process them all at once.

The demo above handles individual pages. If you have a full collection — a box of family letters, a research archive, hundreds of parish register images — the full Transkribus platform lets you upload them all, run the AI on every page, then search, edit, and export the results in bulk.
Upload entire folders — hundreds or thousands of pages at once
Full-text search across everything you have transcribed
Built-in editor to correct results and compare side-by-side with the original
Export as TXT, DOCX, PDF, TEI-XML, or PAGE XML
Train a custom AI model on your specific writer's handwriting for higher accuracy
Open the full platform
Batch processing old handwritten documents in Transkribus

Looking for a specific script or language?

Each page is tailored to a particular handwriting style — with the right AI model, tips, and the demo ready to go.

Old German Scripts

Kurrent, Sütterlin, and Fraktur — the scripts found in German-speaking archives from the 1500s to the 1940s.

Read more

Cursive Handwriting

Convert old or modern cursive writing into readable text. Any language, any style.

Read more

Early Modern Scripts

Secretary hand, Kurrent, Bastarda, humanist minuscule — scripts from the 1500s to the 1800s.

Read more

Medieval Manuscripts

Gothic textura, Caroline minuscule, Beneventan, and other medieval book hands.

Read more

French Handwriting

Ronde, bâtarde, coulée, and French administrative scripts from the 1500s on.

Read more

Italian Handwriting

Humanist hand, notarial scripts, and Italian administrative handwriting.

Read more

Dutch Handwriting

VOC-era scripts, Gothic cursive, and Dutch administrative handwriting from the 1500s on.

Read more

Latin Manuscripts

Church records, legal documents, and scholarly texts in Latin from any century.

Read more

Used by 500,000+ people and leading institutions

Your documents stay private.

Transkribus is built and hosted in Europe by a cooperative — not a Big Tech company. Your documents are processed securely and never used to train AI without your consent.

Your data stays yours

Documents and transcriptions belong to you. Delete anytime.

Processed in Europe

All documents are processed on our own servers in Austria. Fully GDPR-compliant.

No hidden costs

50 free credits every month. No credit card required to start.

Frequently asked questions about reading old handwriting

Transkribus reads handwriting from virtually any century and language. Common examples include German Kurrent and Sütterlin, English and Dutch secretary hand, French ronde and bâtarde, Spanish Procesal, Italian humanist cursive, and Latin church hands. There are over 300 public AI models available, each trained on a specific script, language, or historical period. You do not need to know which script your document uses — upload it and the AI will handle it.

The demo on this page is completely free — no account needed. For larger projects, create a free account to get 50 credits every month (enough for about 50 pages). No credit card required. Paid plans start at a few euros per month for people who need more volume.

Phone photos work fine. Take a clear, well-lit photo of the document with your smartphone, then drag it into the upload box. No special equipment needed. Scans also work — JPG, PNG, and PDF are all supported.

On clean scans with legible handwriting, the AI typically achieves 95–99% character accuracy. Difficult handwriting, faded ink, or damaged paper may produce lower accuracy, but still far better than general-purpose OCR or AI chatbots. You can also train a custom AI model on your specific documents to push accuracy higher.

General-purpose AI tools were trained on modern text and images from the web. They have no training data for historical handwriting — Kurrent, secretary hand, old cursive, and similar scripts are simply not in their training set. Transkribus uses specialised Handwritten Text Recognition (HTR) models trained on over 30 million words from real archival documents. That is the difference.

The AI handles faded ink, foxing, bleed-through, water damage, and low-contrast scans. Results may not be perfect on heavily damaged areas, but Transkribus produces readable output on documents that would be extremely difficult to decipher by hand. The built-in editor lets you compare the transcription side-by-side with the original and correct any errors.

Over 100 languages, from major European languages (English, German, French, Spanish, Italian, Dutch, Portuguese, Swedish, Danish, Norwegian, Polish, Czech, Hungarian) to Latin, Hebrew, Arabic, and many more. The public model catalogue grows regularly as researchers and institutions contribute new models.

Ready to read your old documents?

Create a free account to process full collections, train custom models, search across your text, and export in any format.

Create free accountSee plans and pricing

50 free credits every month – No credit card required

200M+Pages read
500K+Users worldwide
300+AI models