How it works
AI trained on five centuries of Portuguese handwriting
Transkribus uses deep learning trained on millions of handwritten pages, from the 1400s to the present. The AI has learned the characteristic forms of old Portuguese scripts – the broad, rounded strokes of Humanística, the tangled abbreviations of Processual notarial hands, the compact letters of Cortesã – and turns them into modern, searchable text.
Reads Humanística, Processual, Cortesã, Encadeada, and modern Portuguese cursive
Handles the heavy abbreviation typical of Portuguese legal and ecclesiastical documents
Works with phone photos, microfilm scans, or high-resolution digitisations
What you can transcribe
Documents from Portugal, Brazil, and the former colonies
Your document might be a parish register from Brazil, a notarial act from Lisbon, or a ship manifest from a trading post in Goa – Portuguese-language records survive on four continents, and almost everything before 1900 is handwritten. Transkribus reads the handwriting so you can focus on the content.
Parish registers (assentos de batismo, casamento, óbito) from Portugal and Brazil
Notarial acts (escrituras) – property sales, wills, dowry contracts
Inquisition records and ecclesiastical court proceedings
Maritime logbooks, ship manifests, and trading-post ledgers
Colonial land grants (cartas de sesmaria) and royal decrees (alvarás, provisões)
Court records (autos) and municipal council minutes (actas de câmara)
Beyond the demo
The full platform for larger collections
The demo above shows you what the AI can do with a single page. The full Transkribus platform lets you upload entire collections, run batch recognition across thousands of pages, train a custom AI model tuned to your specific handwriting, and search the full text of every transcribed document.
Train custom models on your collection's specific handwriting style
Full-text search across all your transcribed documents at once
Export as TXT, DOCX, PDF, TEI-XML, or PAGE XML
Collaborate with colleagues and share collections across institutions
Common sources
Old Portuguese documents you will encounter
If you're researching Portuguese or Brazilian ancestors, your records may sit in the Torre do Tombo in Lisbon, the Arquivo Nacional in Rio de Janeiro, or a local parish or district archive – and much of it is digitised and free to view online. These are the document types you'll run into most often:
Parish registers (registos paroquiais) – the primary source for genealogy before civil registration
Testamentos and inventários – wills and estate inventories listing heirs, property, and enslaved persons
Livros de notas – notary books recording contracts, land transfers, and legal agreements
Correspondence and petitions to the Crown (requerimentos)
African and Asian colonial records – trade permits, census rolls, missionary reports
Brazilian colonial-period documents – sesmaria grants, Câmara records, Inquisition denunciations
Which script is it?
Found a Portuguese document you can't read? Here's what you're probably looking at.
The script depends on when your document was written and who wrote it – a notary, a priest, or a family member. These are the old Portuguese scripts you're most likely to see:
Humanística – round, fairly clear letters, found in books and formal documents from the 1500s onward. If you can make out most words, it's probably this.
Cortesã – a compact, angular hand in the oldest documents, from the 1400s and 1500s
Processual (Processada) – the notarial and legal script, notoriously hard to read: extreme abbreviation and strokes that run together. Common in escrituras and court files.
Encadeada – an even denser form of Processual where words are chained together without lifting the pen
Modern cursive – from the 1700s onward, handwriting gradually simplified. Later parish registers and personal letters are much easier to read.
The technology
How does Portuguese handwriting recognition work?
Handwritten Text Recognition (HTR) uses deep learning neural networks to convert images of handwriting into digital, searchable text. Unlike OCR for printed books, HTR handles the variation inherent in handwriting – different letter shapes, connected strokes, personal abbreviation habits, and ink bleed that vary from one scribe to the next and from one century to another.
Neural networks trained on large volumes of annotated Portuguese manuscripts
Layout analysis identifies text regions, lines, and marginal notes automatically
Character-level recognition handles the heavy ligatures and abbreviations of Processual scripts
Language modelling uses Portuguese word patterns to improve accuracy on ambiguous letter forms
AI Models for Portuguese Handwriting
Browse public models trained on historical Portuguese documents – Humanística, Processual, colonial-era scripts, and more.
Ready to read old Portuguese handwriting?
Create a free account to process unlimited documents, train custom models, and unlock the full platform. Or try our free old handwriting translator right now – no account needed.
50 free credits every month – No credit card required
200M+Pages processed
500K+Users worldwide
300+Public AI models