OCR vs untxt.: what is the difference?
OCR reads the text off a page. That is one step inside untxt. and not the product itself. untxt. takes the pile OCR would hand back as raw text and turns it into sorted, classified, reviewable records: it knows what each document is, pulls the right fields, maps them, and flags what is uncertain. OCR gives you characters; untxt. gives you bookkeeping data.
What OCR does, and where it stops
OCR turns an image of text into machine-readable text. untxt.'s own text reading is highly accurate, well past the basic OCR most people picture. But even clean text off a page is not a record: it does not know an invoice from a statement, which number is the total, which line is tax, or whether two files are the same bill. Reading the characters, however well, is where plain OCR stops, and where untxt. starts.
What untxt. produces instead
untxt. takes the pile and hands back records: it identifies each document, separates a bundled file into its real documents, pulls the fields and line items, proposes a chart-of-accounts mapping, flags low-confidence values, and gives you a review queue. The output is something you review and post, not text you still have to interpret.
When plain OCR is enough
If your only need is to lift text out of clean, single, searchable PDFs, an OCR tool may be all you need. The difference shows up with mess: mixed PDFs, photos, varied vendors, jumbled document types. That is where reading the text is the easy part and the pile is the job.
Why "is it just OCR" is the wrong question
The hard part of intake was never reading a clean line of text. It is everything around it: what the document is, what belongs together, what is uncertain, where it posts. untxt. is built for that part. OCR is a component, not the competitor.