wondarchive
EN
Log in Create my free account

wondarchive › General AI and archives

Can ChatGPT or another AI assistant transcribe your archives?

For a page or two, yes: an assistant like ChatGPT reads a sharp photo and gives you its text. For a collection of several thousand photos, it lacks what makes a historian’s work: the order of the box, the shelfmark and folio, search across the whole collection, consistent translation and a place where everything stays.

What does a general AI assistant do well?

It reads a sharp photo of a typewritten, printed or handwritten page, gives you the text, translates it, explains an archaic word or an abbreviation. To decipher a difficult line or understand a single document, it is a good first move.

Where does it fall short for a whole collection?

  • Photos go in a few at a time, conversation by conversation; nothing is filed by repository, shelfmark and folio.
  • You can’t search for a name across five thousand pages: you have to remember which conversation it came up in.
  • Translation changes from one conversation to the next: the same technical term can end up with several translations.
  • An illegible word can be filled in with a plausible one without being flagged, and a near-blank page can be given a text.
  • The text isn’t shown next to the photo, page after page, for checking.
  • Whether images are used to train models depends on the service, the plan and the account settings.

What more does wondarchive do?

NeedGeneral assistantwondarchive
Uploading two thousand photosa few images per messagethe whole folder, subfolders included
Turning pages the right way upto do beforehandautomatic, on your computer
Handwritingread, page by pageread by extra-strength reread, from the full-resolution original
Filing—repository, shelfmark, folio, as in the archive
Searchingwithin one conversationacross the whole collection, names with their variants, dates
Translatingon requestevery page, with the context of neighbouring pages and the collection’s glossary
Blank versomay be given a textrecognised, no text invented
Citing—collection, shelfmark and stable folio
Your imagesdepending on the service’s termsstored in France, with OVHcloud, read in the EU, no training

Do you have to choose one or the other?

No. An assistant answers a question about a page well; wondarchive keeps the whole collection read, filed and searchable, from one archive trip to the next and all the way to your viva. wondarchive is an online service that transcribes, translates and indexes archive photos taken in the reading room, for historians, PhD students, archivists and genealogists.

All the featuresPricingCreate my free account

The account takes a minute to open, a hundred pages free for a year.

Can ChatGPT read old handwriting?

It often reads a sharp photo of handwriting; the result needs checking against the image, because an illegible word can be replaced by a plausible one. wondarchive shows the text next to the photo for that check, page after page.

How many photos can you send to an AI assistant?

A limited number per message, depending on the assistant and the plan; a thesis collection often runs to several thousand. wondarchive takes the whole folder in one upload.

Are my photos sent to an assistant used to train it?

That depends on the service, the plan and the account settings: read its terms. In wondarchive, uploaded documents train no model; they are stored in France and read in the EU.

Does wondarchive use AI?

Yes: every page is read by wondarchive’s AI, from the full-resolution original, and handwriting goes through an extra-strength reread with our most precise model. The models are called from our servers, in the European Union.

See it on your own pages

A hundred pages transcribed and translated, for a year, no card required.

Create my free account See pricing