How to transcribe and translate Latin archival documents

Can you automatically transcribe a Latin archival document?

Yes: automatic transcription of a Latin document is possible, whether it is printed (OCR) or handwritten (HTR, handwritten text recognition). At wondarchive, Latin is one of the European languages transcribed in the original language and then translated into yours. On a demonstration page, printed Latin is read almost perfectly: under 1% character errors on a printed text from 1777. A handwritten Latin page goes through enhanced review, using our most precise model, which reads cursive hands in the Latin alphabet. The text appears beside the photo so you can check a name or a date at a glance.

wondarchive is an online service that transcribes, translates and indexes photos of archives taken in the reading room (manuscripts, typescripts, printed material), for historians, doctoral students, archivists and genealogists. The detail of what is read, page type by page type, is on what wondarchive reads.

Which Latin documents are read well, and which need checking?

Printed Latin documents are read best; well-written modern Latin manuscripts are read almost without error; older registers and deeds need a check of names, dates and abbreviations. The quality of a Latin transcription depends first on the sharpness of the photo, then on the type of script. Here is how the most common cases in a Latin collection break down.

• Printed Latin (books, dissertations, ordinances, periodicals): near-perfect reading on a demonstration page (0.4% character errors on a printed text from 1777); figures are put back in their place in the text.
• Latin manuscripts in a legible cursive hand (parish registers, deeds, correspondence): enhanced review; the Latin formulae are accurate, but proper names, especially those from another language such as German surnames in a register, and dates should be checked against the photo.
• Heavily abbreviated medieval hands (charters, cartularies): the structure and formulae are read, but a name, a place or a year may be misread: treat it as a reading aid and check line by line. Try the 100 free pages before committing a whole collection.
• Blank versos: recognised as such; no text is invented.

How do you transcribe and translate a Latin collection with wondarchive, step by step?

To transcribe and translate a Latin collection with wondarchive, simply upload the folder of photos as it is, then read, check and search: there is nothing to prepare. Transcription and translation run on the server, in batches, within minutes or hours of the upload, even with the tab closed.

1. Upload the folder for your research trip, subfolders included; the original JPEGs are kept intact.
2. Let the pages be straightened: the browser decides the orientation of each page and only asks you about doubtful cases.
3. Read the transcription in Latin, photo on the left, text on the right; the language, date and proper names of each page are picked out.
4. Read the translation with one click; each page is translated with the end of the previous page and the start of the next as context.
5. Fix the vocabulary of the collection in the glossary (canonical, legal and place terms) for a consistent translation from start to finish.
6. Correct whatever needs it: the machine-read version is kept and each correction is dated (correcting the transcription).
7. Organise: location, shelfmarks, folios, then export.

Are Latin abbreviations transcribed correctly?

Latin abbreviations are not expanded in a guaranteed, uniform way: in any automatic transcription of a Latin manuscript, whatever the tool, they are the first thing to check. Verify them against the photo, displayed next to the text, and correct them to the form you have chosen for your edition. To resolve a difficult abbreviation, Adriano Cappelli's Lexicon abbreviaturarum remains the standard reference for palaeographers.

A simple rule for a thesis: decide from the start whether you will expand abbreviations or keep them as written. Record this choice in your methodology, and apply the same convention to all your corrections.

Is automatic Latin translation reliable enough for a thesis or family history?

Automatic Latin translation is reliable for understanding and sifting a collection and spotting the useful items; a name should be confirmed on the photo, since that is where the reading goes wrong most often. A citation in a thesis should rest on the verified Latin text, with your own translation if needed. At wondarchive, the translation follows the collection page by page, with the context of neighbouring pages and the collection glossary, so the same term is not translated differently from one page to the next.

For a genealogist, the translation is often enough to identify a baptism, a marriage or a burial; the name, date and godparents are then checked on the photo. The criteria for judging a transcription and checking it quickly are detailed in how reliable is automatic transcription.

Which tool should you choose to transcribe Latin archives?

For photos of Latin archives to transcribe and translate without installing anything, wondarchive is a good fit. To train a model on one specific hand, Transkribus or eScriptorium meet a different need, and a general-purpose AI assistant suits a single page. The right choice depends on volume, goal (reading, editing, training) and available technical resources. State as of October 2026.

The detailed comparison: Transkribus alternative.

Follow-up questions

Are my Latin archive photos used to train a model?

No. At wondarchive, images are stored in France, read in the EU and are not used for any training. The whole collection can be exported if you leave.

Can I export the Latin transcription?

Yes. A page, a shelfmark or a search can be exported as PDF (exporting pages).

Can I find a Latinised name, such as Joannes for John?

Yes, provided you search for the Latin form as it is written. Full-text search covers the whole collection and indexes proper names with their variants. For an inflected name (Joannes, Joannis, Joanni), run the search on several forms so as not to miss anything.

What about Latin parish registers?

Latin parish registers are handled like any manuscript: enhanced review, then checking the name and date against the photo. The case is detailed in reading parish registers and civil records.

How can I tell whether my Latin hand is read well?

The simplest way is to try it: the free account offers 100 pages transcribed, translated and kept for one year, with no credit card. Choose the most difficult pages in your collection.

To judge on your own Latin pages: open a free account (100 free pages) or browse the demo.