Turning a PDF Into Audio When the PDF Is a Scan

If you upload a scanned PDF here, it gets rejected. There is no OCR on this platform, and a scan is not text — it's a picture of a page. A voice generator can only read characters it can find in the file, and a scanned page has none, no matter how sharp the image looks on screen.

That is the limitation, stated plainly, and it is not going to change by trying a different file name or a smaller file. What follows is how to check whether your PDF is actually a scan, how to fix it in the tools you probably already have, and what to watch for once you've fixed it and are ready to generate the audio.

You can find out in about five seconds without opening any special software. Open the PDF in any reader and try to click-and-drag to select a line of text. If a text cursor appears and a sentence highlights, the file has a real text layer and you're fine. If nothing highlights, or your cursor turns into a little rectangle for drawing a selection box over the image, it's a scan.

A second check that catches the same thing: press Ctrl+F (or Cmd+F) and search for a word you can see printed on the page. A text-based PDF finds it instantly. A scanned PDF finds nothing, because there is nothing to search — just pixels.

The fix is OCR, and you almost certainly already have access to it somewhere. OCR — optical character recognition — reads the shapes of letters in the scanned image and turns them into actual, selectable text. We don't do this step for you, but three places usually already do.

Once the OCR pass has run, save the result as TXT or DOCX rather than re-saving as a PDF. A plain text or Word file guarantees the text layer travels cleanly and strips out any leftover image data you don't need.

OCR is good at recognizing letters and bad at understanding page layout, so it hands you back two specific problems almost every time. Fix both before you generate audio, not after — they're a five-minute find-and-replace job now and an annoying re-listen later.

Once a document has a real text layer, four formats are accepted for turning it into audio: EPUB, TXT, PDF (text-based, not scanned), and DOCX. There's no MOBI support and no audiobook file (M4B) output — you get MP3, and WAV if you're on Pro or above.

Generation is priced in points, and points work out to roughly one point per second of finished audio. The exact cost is calculated from character count rather than duration, so treat the per-second figure as a useful estimate rather than an exact rule — a page with a lot of long words will cost a few more points than the seconds of audio suggest, and a page of short, punchy sentences will cost a few less.

That difference is expected, not a bug. It's also a reason to fix the OCR problems above before you generate: a document still cluttered with running headers and hyphen breaks costs you points reading text nobody wanted narrated in the first place.

With a real text layer in hand, you have the full range of document-to-audio tools available. A textbook chapter, a scanned government form you converted to DOCX, or an old EPUB you've had sitting around all go through the same path from here.

Pick a voice that fits how the document will actually be used. A dense technical manual benefits from a clear, unhurried reader; a report someone will listen to on a commute can carry a bit more warmth. The six voices below cover both ends of that range — press play on any of them before you commit.