Some of a PDF converts to audio cleanly. Some of it disappears without warning

This page is about which parts survive. WeSolve+ is best for someone who wants the argument of a chapter in their ears rather than a machine reading forty pages at them.

First the file has to give up its text

A PDF stores instructions for drawing a page, not a stream of sentences, so a document either carries a text layer or it does not. If it was exported from a word processor it usually does. If it is a scan or a photograph of a page, the characters have to be recovered by optical character recognition, which introduces its own errors and is silent about them. The quick test takes five seconds: try to select a sentence with your cursor. If you cannot, there is no text layer and everything downstream is a reconstruction.

Then the reading order has to be recovered

Two columns, a running header, a page number, a figure caption and a footnote are all separate flows on the same page, and nothing in the file necessarily says which order a human reads them in. Working that out is layout analysis, and when it fails you hear it immediately: a sentence stops mid clause and a page number is read aloud in the middle of it. Files exported with structure, what the standard calls a tagged PDF, carry the reading order explicitly and convert far better, which is also why they work with screen readers.

What is lost silently, and this is the important part

Three things vanish without any signal. A table becomes a stream of numbers with no structure, so a row and a column sound identical. A formula becomes either nothing or a mangled phrase, unless the document carries it as MathML rather than as an image. And a figure becomes its caption, which usually says see figure four. Because none of these produce an error, a badly converted document sounds like a well converted one, and you only find out in an exam. The rule that follows is to read anything visual and listen to anything argued.

Accessibility work already solved most of this

Everything above has been studied for decades under accessibility rather than under study tools, because a blind reader has needed a document read aloud since long before anybody wanted a podcast of their lecture notes. That literature is where the answers are: tagged structure, a text alternative for every image, tables marked up as tables. It also gives you a free implementation, since every current operating system will speak selected text through built in speech synthesis, and a screen reader handles a well tagged document better than most conversion services.

A forty page document should not become forty pages of audio

Reading a chapter aloud in full takes about four times as long as reading it, and produces something nobody finishes. The useful artefact is short: the argument, the two or three anchor facts, and the thing you keep forgetting. That is what the tool above builds, by keeping roughly one sentence in five of whatever you paste, in your own wording. Two minutes of audio holds about three hundred words, and a recap you can get through on the walk to a lecture is worth more than a complete recording you never open. There is more on when listening beats reading on the podcast generator page.

The recap the app makes from the file itself

Upload the PDF, photograph the page or take a camera shot and WeSolve+ reads it and produces a spoken recap of roughly one minute, along with an explained quiz and a card deck from the same file. The recap is deliberately short for the reason given above, and it is built from the document rather than from your extracted text, so you do not have to do the pulling out yourself. On the free plan there are two recaps a week. At wesolveapp.com in any browser and on the App Store for iPhone and iPad. Figures on pricing, and the feature on audio recap.

Last updated: 2026-08-15