Convert TIFF to TXT

Almost nothing arrives as TIFF by accident. It comes from a scanner, a document-management system, a records department, an archive or a fax gateway — and it is usually paperwork: contracts, correspondence, claims, case files, medical records, minutes.

That changes what the job is. Nobody is trying to copy a paragraph off a picture; somebody is trying to make a body of documents searchable, or to pull quotable text out of an archive, or to get figures out of a stack of forms. The volume is higher and the accuracy matters more.

Everything that makes a phone photograph hard is absent. The page is flat against glass, so there is no perspective. The lighting is even, so there is no shadow across the text. The resolution is known and consistent. And a scanner is focused at exactly the right distance by construction.

The practical result is that clean printed type scanned at 300 dpi comes back close to perfect. That is a genuinely higher expectation than the photograph pages on this site can offer, and it is why bulk recognition work is done from scans rather than from pictures.

Resolution first. Below about 200 dpi there are not enough pixels per character and accuracy falls away sharply; 300 dpi is the working standard and 600 is worth it for small type or poor originals.

Then the original itself. Faxes are the worst common case — low resolution by design, often several generations of copying deep. Carbon copies, thermal paper that has faded, documents with stamps or handwriting over the print, and typewriter faces with broken characters all produce errors. So does a scan that came out skewed, which is worth correcting before recognition rather than after.

A scanner will frequently produce a single TIFF containing every page of a document. Only the first image in the file is read, and there is no page control to move past it, so a ten-page scan returns the text of page one and nothing else. Nothing announces the other nine — the output simply stops.

On documents scanned a page to a file that costs nothing. On a bundle it is the point at which the honest answer is that the pages have to be separated before they get here, by the scanner or the records system that wrote the TIFF. This page is for the documents you actually need to read now, not for digitising a filing room.

Recognition matches shapes against letterforms. It does not understand the document and it cannot tell you where it was unsure, so what comes back is fluent, complete-looking text that may contain mistakes nobody flagged.

In records work that matters more than usual. A misread digit in a policy number, a date, an amount or a case reference is exactly the error that survives a quick read and causes a real problem later. Anything going into a system, a bundle or a citation should be checked against the page it came from.

For an archive, plain text throws away the thing that makes a scan worth keeping: the image of the page, which is the record. A searchable PDF keeps that image and puts a text layer behind it, so the document still looks like itself, still prints like itself, and can be searched and copied from.

Plain text is the right output when you want the words as words — to quote, to index, to feed into something, to count. It is worth deciding which of the two you actually need before converting a hundred files, because the two workflows diverge immediately.

Recognition happens inside the browser tab on your own processor. Nothing is uploaded, so there is no server copy, no retention period to ask about, no account and no daily allowance.

Given what is typically in these files — signed contracts, patient records, court filings, insurance claims — that is the substantive difference from a hosted service rather than a feature line. It also removes the awkward question of whether sending a client’s documents to a third party needed anybody’s permission.