A natural-sounding voice has to be generated somewhere, and where that happens is not a detail. It decides whether the text of your book leaves your device at all. This is worth two minutes of your attention, and then you can go back to not thinking about it, which is rather the point.
The common approach is simple enough. The app sends the text of a page to a remote server, a neural voice model running there turns it into audio, and the audio is streamed back to you. It is a sensible design: the heavy model lives on hardware far more powerful than a phone.
It also means the sentences you are reading travel to a company's servers to be processed. For a recipe, nobody cares. For a book about a diagnosis, a divorce, a faith or a job you are quietly thinking of leaving, some people care rather a lot.
PhonoLeaf's natural voice is generated by a neural model that runs on your device rather than on a server. The text of your book never has to be sent anywhere, because nothing anywhere else is involved in making the sound.
The practical version: no backend of ours holds your books. Not a well-secured one, not a well-intentioned one. Your books have nowhere to be sent, which is a stronger guarantee than a promise not to send them.
It means the app works on a plane and in a tunnel, because there was never a round trip to lose. It means the voice does not quietly get worse when your signal dips, because there is no lower-bandwidth fallback to drop to.
And it means the honest answer to "what do you do with my reading data" is that we never receive any. That is less a policy than an architecture.
Once a voice is downloaded and a book is saved for offline listening, yes. Both speech generation and playback work with no connection at all.
PhonoLeaf uses published, open neural voice models, not a model trained on your reading. They are downloaded once and run entirely on your device.
No. Voice generation happens on your device, so the text of a book never leaves it and there is no reading history on our side to collect. The only thing the app downloads from us is the voice file itself, which is identical for every reader.
No. It is read-only, and you can revoke it at any time from your Google account settings, or by signing out in the app, which revokes the grant too.
Seven days free, and your reading stays where it is.