Read a PDF Out Loud
This page reads a PDF out loud using the Web Speech API voices already installed with your browser and operating system — no account, no upload, no word limit and no cost. It extracts each page's text, splits it into sentences, and speaks them in order with page, speed and voice controls, showing sentence-by-sentence progress and advancing to the next page automatically. Everything runs on your device.
How it works
Two on-device pipelines are chained together. The first is text extraction: pdf.js parses the document and returns each page's text as positioned runs, which the tool assembles into reading order by their coordinates on the page and then joins into a continuous passage. That passage is segmented into sentences — splitting on terminal punctuation while stepping around abbreviations and decimal numbers — because sentences are the natural unit for speech: they give the voice sensible places to breathe, give you a visible progress marker, and mean a pause or page jump never restarts a huge block of text.
The second pipeline is the browser's own speech synthesizer, exposed to pages as the Web Speech API (speechSynthesis). Each sentence becomes an utterance queued with your chosen voice and rate; when the engine signals the end of one utterance, the tool highlights and queues the next, and when a page's last sentence finishes it loads the next page's text and keeps going. The voice list is whatever your operating system ships — different on Windows, macOS, Android and iOS — which is why quality varies by device and why the menu on your machine may look different from a friend's. Locally installed voices synthesize entirely offline; no text is sent to any service.
One quirk is worth explaining because it looks like a malfunction: browsers refuse to let a page produce audio before you have interacted with it, an autoplay protection that applies to speech synthesis just as it does to video. That is why reading starts from the Play button rather than the moment your file loads — the click is the permission. And since the whole approach begins with extractable text, a scanned PDF is silent until it has been through OCR; the invisible text layer OCR adds is exactly what the reader consumes.
Illustrative example: a 32-page report on the commute
Illustrative scenario, not a recorded benchmark. File sizes, timings and results depend on the input and device. Try the reproducible samples.
A 32-page policy report of about 9,400 words was loaded before a train journey. At the default 1.0× rate — roughly 170–180 words per minute for most system voices — the full document runs about 54 minutes; at 1.5× it fits the 36-minute ride. The listener pressed Play (the required first click), picked a preferred system voice from the 14 the laptop offered, and jumped straight to page 5 to skip the title page and contents. Sentence highlighting tracked the reading — 22 to 30 sentences per page — and each page rolled into the next automatically. Airplane mode on, the reading continued unaffected: the voice lives on the device.
Frequently asked questions
Is reading a PDF aloud here really free and offline?
Yes. The voices belong to your operating system and browser, not to a paid cloud service, so there is no word limit, no account and no upload. Once this page has loaded, speech works without an internet connection for locally installed voices.
Why does nothing play until I press the Play button?
Browsers block pages from producing sound until you interact with them — an autoplay protection, not a bug. The first click on Play counts as that interaction and unlocks the speech engine; after that, pause, resume and page changes work normally.
Why do the voices sound different on my phone and my laptop?
The tool uses whatever voices your operating system and browser ship, and every platform bundles different ones. Voice quality and the languages on offer therefore vary by device; the voice menu lists exactly what yours provides.
Can it read a scanned PDF out loud?
Not directly, because a scan contains images of text rather than extractable text. Run the file through the OCR tool first; the recognized text layer it adds is exactly what this reader needs.
Can I change the reading speed or skip to another page?
Yes. The rate control adjusts speech from half speed up to double speed, and the page control jumps anywhere in the document. When a page's last sentence finishes, reading advances to the next page automatically.
Is my PDF uploaded to a text-to-speech service?
No. Text extraction happens in your browser with pdf.js and speech is generated on your device by the Web Speech API. The document's content is never transmitted anywhere.