Documents, scans and interviews as searchable text.
An object comes with letters, index cards, expert reports, interviews and literature. Moliri reads the text from scans and recordings, makes it searchable and keeps the literature in a library of its own, on the same record as everything else.
We set up transcription and server-side text recognition on your instance.
Text from scans and photos
In the file window, Moliri recognises the text of a scan or a photographed document, Fraktur included. The recognised words lie as frames over the image, and uncertain passages are marked.
The reading happens in the browser, with the desktop app Transcribe on your computer, through the AI service or on the server. In the browser and with Transcribe, the image does not leave the computer. Each person chooses the default in their profile, and it can be changed for a single run.
The AI service is an image model that your instance addresses through its AI account, on your own server or with a provider. It also reads Latin handwriting, line by line; it does not read Kurrent reliably. For this the image goes to that service, and the instance can switch it off.
You correct in the text alongside. Corrections are kept wherever the line is found again when the text is recognised anew, no matter where the reading happens, and they appear in the history. The text can be downloaded as a text file.
In the handbook: Recognised text · Text recognition in the profile
Text recognition in the file window
- WhereIn the browser, with Transcribe on this computer, through the AI service or on the server
- LanguagesGerman, German and English, English, German + Fraktur
- LayoutPage, lines as on an index card, a label, writing on objects
- HandwritingLatin script through the AI service, Kurrent not reliably
- ResultText with word frames over the image, with lines when the AI service reads
- CorrectionIn the editor, with search, recorded in the history
Interviews and recordings as text
Audio and video files become a transcript with time marks and speakers. Speakers can be named and merged, corrections are marked, and important passages are kept as clips with a title and keywords.
You choose for each recording whether Moliri transcribes it straight away or overnight. Rules by keyword do that for new files. The transcript goes out as text, as subtitles or in the noScribe format, which Word and MAXQDA can process further.
In the handbook: Transcript · Rules and night window
Transcript of a recording
- RecordingsAudio and video from the file library
- SpeakersRecognised, can be named and merged
- ClipsPassages with a title and keywords
- OutputText, subtitles (VTT), HTML for noScribe
Full-text search in documents and recordings
Recognised text and transcripts are part of the full-text search, as a search area of their own that you switch on and off. A hit names the file, for a recording also the passage, and opens the transcript right there.
That way, a search for a name also finds the letter that mentions it and the interview in which it comes up.
Search areas of the full-text search
- FieldsThe values of the objects
- MetadataComments and sources at the field
- EventsExhibitions, loans, acquisition and more
- Image metadataDescriptions of the files
- Transcripts and OCRRecognised text and transcripts
Index cards and lists as objects
A pile of typed index cards, inventory lists or old catalogues becomes objects. Moliri recognises the text, splits it by document, page or entry and maps the details to your fields, by their labels, by position or by your earlier decisions.
Headers and page numbers stay out, and illustrations go into the object's image field. All pages are shown one below the other, with the tools beside them: a click on a text gives it a field, a frame excludes a letterhead on all pages, a line separates two objects on one page. On the right you see the objects as index cards, as a table or field by field. The result runs through the familiar Excel import, with preview and undo.
Excel import · In the handbook: OCR import
Taking over documents and scans
- FilesMany PDFs and images at once
- Text recognitionIn the browser, with Transcribe, through the AI service or on the server
- SplittingOne object per document, per page or several per page, detected after reading where it is clear
- CheckingObject boundaries, fields and illustrations right on the page, preview of the finished objects
- MappingBy label, position, order, content or earlier decisions, with AI if you wish; a text from the title page for all objects
- NextAs an Excel file or straight into the import
Literature with references and citation style
Literature lives in a library of its own and is linked from the objects. A new title is created from an ISBN, a DOI or a title: Moliri fetches the details from the Deutsche Nationalbibliothek (German National Library), from K10plus, Crossref or Wikidata, and links the authors through the GND. This also works from the field on the object, without leaving the object. If the title is already in the library, Moliri tells you before a second one is created.
The Citation list view outputs the literature as a bibliography, in the citation style of your choice from the catalogue of the Citation Style Language. The same style applies in the Word report and the Excel export.
Literature library
- New from a referenceISBN, DOI or title
- SourcesDeutsche Nationalbibliothek, K10plus, Crossref, Wikidata
- AuthorsLinked through the GND
- Citation stylesCSL, selectable per view, report and export
Dictate instead of typing
You can dictate into any text field, using the microphone at the field or a keyboard shortcut, for a condition report in the store, say. The recognition runs on your own computer with Transcribe by Moliri, directly in the browser, or through a service of your choice. In the browser, a one-off download of the speech model is enough, with no key and no extra program; after that, your speech stays on the computer. In no case does the recording pass through Moliri's server.
Transcribe by Moliri: the free app for Windows that also transcribes interviews and reads scans on your computer.
All features at a glance
- Objects
Find, organise, change
Read more - Data model
Catalogue the way your institution catalogues
Read more - Checks and data care
Keeping the data clean
Read more - Documentation and research
Documents as searchable text
This page - Conservation
Findings belong to the object
Read more - Files
One file library for all projects
Read more - Exhibitions and loans
From loan request to the finished wall
Read more - Standards and publication
Map once, deliver anywhere
Read more - Reports and export
From the database to the document
Read more - Setup and administration
Your institution carries the instance on
Read more
Bring a document
A scanned index card, a letter in Fraktur or an interview. In the demo we turn it into text, search it and attach it to the object.