The problem
Why old family letters are so hard to read
You found a bundle of letters in a drawer, an attic box, or a family estate. You know they matter – but you can't read them. The handwriting is faded, the ink has browned, and the letterforms look nothing like what you learned in school. Older letters often use scripts that changed dramatically over two centuries, from the ornate flourishes of copperplate to the compressed loops of wartime correspondence.
Faded iron gall ink – the brown or sepia tone means the ink is literally eating through the paper over time
Water damage, foxing spots, and fold-line wear where the paper was creased for decades
Cross-writing – a common 18th and 19th-century practice of writing over existing text at 90 degrees to save paper and postage
Unfamiliar script styles – copperplate capitals, Spencerian flourishes, and Palmer method shortcuts look alien to modern eyes
No standardised spelling – the same word or name can appear three different ways in a single letter
The solution
AI that reads centuries of English handwriting
Transkribus uses handwritten text recognition (HTR) – AI trained on millions of historical handwriting samples. Unlike general-purpose AI tools, which handle modern handwriting reasonably well but struggle with older scripts, Transkribus was built for exactly this kind of document: faded ink on old paper, unfamiliar letterforms, and damaged pages. Upload a photo of your letter, and the AI returns readable text in seconds – with confidence scores on every line so you know which words to double-check.
300+ public models – including models trained on English-language handwriting from the 1700s to the 1900s
Handles faded ink, water damage, bleed-through, and low-contrast documents
Reads copperplate, Spencerian, Palmer method, and dozens of other historical scripts
Works from a smartphone photo – no scanner needed
Export as plain text, PDF, or structured XML
Identification guide
What script is that? Dating your family letter
If you can roughly identify the handwriting style, you can date your letter and choose the right AI model. English-language handwriting went through three major phases between the 1700s and the mid-20th century – each with distinctive letterforms that the AI has been trained to recognise.
Copperplate (c. 1700 – 1850) – formal, elegant, written with a pointed pen that produces thick downstrokes and thin upstrokes. Common in legal documents, formal correspondence, and early American letters
Spencerian (c. 1850 – 1920) – lighter, more delicate, with distinctive oval forms and flowing connectors. The standard business and personal hand of the Civil War and Gilded Age
Palmer method (1890s – 1960s) – simplified, faster, designed for commercial use. Muscular arm movement replaced finger motion, producing rounder, more uniform letters
Cross-writing – not a script but a paper-saving technique common when postage was charged per sheet. Text written in one direction, then the page rotated 90 degrees and written over again
Before you hire a service
Before you hire a transcription service
If you searched for a professional transcription service, you're not alone – many people assume a human expert is the only option. Professional transcribers charge per page and can take weeks to return results. Before committing to that cost, try the AI for free. Transkribus handles most legible handwriting well, and you'll know within seconds whether the result is usable. For legally critical documents – wills, deeds, or anything submitted to a court – a professional review still makes sense. But for family letters, the AI is often all you need.
Try the AI first – 50 free credits every month, no credit card required
Results in seconds, not weeks
Confidence scores help you spot the words that need a human eye
For legal or archival-quality work, use the AI draft as a starting point for professional review
AI models
300+ AI models for historical handwriting
Transkribus's public model catalog includes models trained by researchers, archivists, and genealogists on real historical collections. Many are optimised for specific time periods, regions, or document types – so you can find a model that matches your letters closely. The community adds new models regularly, and you can train your own if your documents use an unusual hand.
English-language models covering 18th, 19th, and 20th-century handwriting
Models for German Kurrent, Sütterlin, French, Dutch, and 100+ other scripts – useful if your family letters include non-English correspondence
Filter by language, century, and script type to find the best match
Train a custom model on your own letters if public models don't cover your specific hand
Frequently asked questions
Ready to transcribe your old family letters?
Create a free account and start reading your family's letters today. 50 free credits every month – no credit card required.
50 free credits every month – No credit card required
200M+Pages processed
500K+Users worldwide
300+Public AI models