PDF to Audiobook — Free, With Natural AI Voices

Reports, papers, and half the books people actually own live in PDF. Converting a PDF to an audiobook shouldn't mean a subscription or a robotic voice — here's the free, open-source way that produces a real listenable series.

How the conversion works

  1. Text extraction: the pipeline pulls the text layer out of your PDF, strips page-number noise, and normalizes paragraphs. (Scanned PDFs without a text layer need OCR first — the tool tells you instead of producing silence.)
  2. Chunking: the text is split into ~10-minute episodes at sentence boundaries, each with a one-line AI summary so the series is navigable.
  3. Synthesis: neural voices (the family behind Edge's Read Aloud) render each episode; an AI-generated intro episode has two hosts discuss what the document argues.
  4. Delivery: everything lands in your private podcast feed and a PWA player — resume positions, speed control, offline caching.

Start listening in minutes

Generation runs about 10x faster than playback: upload a PDF from your phone's browser and the first episode is ready before you've laced your shoes; a 300-page PDF finishes in the background while you listen.

Why not a "PDF reader" app?

Read-aloud apps play one document while the screen is on, and lose your place when it locks. A podcast-feed pipeline gives you lock-screen controls, per-episode resume, and a queue — the difference between "text to speech" and an actual audiobook.

listenwhilerunning is open source (AGPL-3.0) and runs on your own free-tier accounts.

Get it on GitHub → Live demo Hosted version waitlist

FAQ

Does it work with scanned PDFs?

Not directly — scanned pages have no text layer. Run OCR first (e.g. with ocrmypdf), then upload the result.

Are equations, tables, and footnotes handled?

They're read as text, which works for prose-heavy documents. Heavily mathematical papers are better skimmed visually and listened to for the prose sections.

What does it cost?

Nothing per book: the pipeline is AGPL open source, storage fits Cloudflare's free tier, and the default neural voices are free for personal use.