Most books never get an audiobook edition, and commercial conversions charge per book. With a modern neural TTS pipeline you can convert your own epubs into serialized audiobooks for free — and start listening minutes after upload.
A raw 10-hour MP3 is unusable on a run. listenwhilerunning splits each book into chapter-sized episodes (~10 minutes each) with per-episode AI summaries, so your player shows a proper series: resume where you left off, skim what each part covers, jump around freely.
Audio generation runs roughly 10x faster than playback. Upload an epub from your phone; the first episode is ready in a few minutes, and you can listen while the rest of the book generates — you'll never catch up to it. A 200-page book completes in about 2–3 hours in the background.
Before episode one, the pipeline generates a bonus episode where two AI hosts discuss the book: what it argues, why it's worth your time, and what questions to keep in mind. It's a surprisingly good way to decide whether to commit ten hours to a book.
The pipeline detects the book's language and picks a matching neural voice automatically. There's also a whole-book translation command: feed it an English epub and listen to it as a Chinese audiobook (or configure any language pair your LLM supports).
Most epub-to-audiobook converters meter you by characters or library size, so a full book — typically 300,000–600,000 characters — effectively requires a paid plan. An open-source pipeline on your own accounts has nobody to bill you:
| listenwhilerunning (this project) | tts.ai | Éist (eist.app) | |
|---|---|---|---|
| Price | $0 — runs on your own free-tier accounts | Free tier 10,000 chars/month; plans $5–$99/mo by character quota | Free with 1 book in library at a time; Premium $4.99/mo |
| Full books without paying | Yes, unlimited | No — one book exceeds the free monthly quota many times over | One at a time (unlimited hours on that book) |
| Open source | Yes, AGPL-3.0 | No (proprietary platform) | No |
| Self-hosted / your storage | Yes — your own Cloudflare R2 bucket, plain MP3s, no DRM | Self-hosted Docker offered from $99/mo | No — iOS/Android app, audio stays on device |
| Delivery | Standard podcast RSS: any podcast app, resume, lock-screen controls | In-app / download | In-app player with word sync and speed-reading mode |
| Extras | AI two-host intro episode, per-episode summaries, whole-book translation | Large voice/model catalog | Fully offline, on-device synthesis (a real privacy plus) |
Pricing and limits checked on the vendors' public pages in July 2026 — see their sites for current details. Both are legitimate tools; the difference is a product you rent vs. a pipeline you own.
epub, txt, and html directly. For mobi/azw3, convert to epub first with Calibre (one click).
It uses Microsoft's neural voices (the same family behind Edge's Read Aloud), which are close to human narration for most prose. For commercial use you can swap in any TTS provider via a small interface.
Yes — the pipeline is AGPL open source and runs on your own Cloudflare free tier; the default neural voices are free for personal use. There is no plan, quota, or per-book fee, because there is no service in the middle.
Converting books you own for personal listening is generally fine. The tool is self-hosted and private by design — your files never leave your own storage.