Sometimes you don't want an app — you want MP3 files. This open-source pipeline converts an epub into a series of ~10-minute MP3s with natural neural voices, stored in your own bucket where you can download, sync, or podcast them.
Each book becomes bookaudio/<id>/1.mp3, 2.mp3, … in your own object storage — 96kbps mono, sentence-boundary splits, roughly 10 minutes per file. They're plain MP3s: copy them to a watch, an old iPod, a car USB stick, anything.
The same files are also published to a private podcast RSS feed with episode titles, AI-generated per-episode summaries, and an intro episode where two AI hosts discuss the book. Subscribe once in any podcast app and you get resume positions and lock-screen controls for free — worth trying before you go file-management mode.
The default voices are Microsoft's neural TTS — the same family as Edge's Read Aloud — which handle long-form prose with natural pacing and are free for personal use. Language is auto-detected per book, and you can pick any voice from the catalog.
Yes — they're regular files in your own Cloudflare R2 bucket with public URLs. No DRM, no app lock-in.
About 5MB per 10-minute episode at 96kbps mono; a full book is typically 300–600MB, and R2's free tier holds 10GB.
PDF works directly; mobi/azw3 convert to epub first with Calibre (one click).