# Voice packs

The code and the rest of the app are licensed under the PolyForm Noncommercial
License 1.0.0 (see `LICENSE`). **The voice packs are different.**

## Recorded voices (all rights reserved)

Any voice pack under `audio/voices/` **other than `griff`, `ana` and `owl`** (for
example `audio/voices/mom/`, `audio/voices/dad/`, or any pack added later) is a
recording of a real person's voice. These recordings are

- copyright the author, **all rights reserved**;
- **not** covered by the PolyForm Noncommercial License or any other license
  in this repository;
- **not** for reuse, redistribution, training, cloning or imitation of the
  speaker's voice, in whole or in part, in any form.

You may play them inside this app as shipped. If you want to redistribute the
app, remove those folders first; the app works without them.

Voice packs you record yourself with the in-app recorder are yours, and stay on
your device unless you export them.

## Griff (`audio/voices/griff/`), the default

Synthetic speech from Fish Speech 1.5 (code Apache 2.0; model weights
CC-BY-NC-SA 4.0, **non-commercial**), cloning 15 seconds of "Barbara Bear"
narrating The Velveteen Rabbit for LibriVox (archive.org `velveteenduet_librivox`,
public domain). The reference clip is in `tools/voices-ref/`. These clips may be
used and shared non-commercially with this app; a commercial build needs a voice
made with a commercially licensed model. The 26 letter-sound clips in this folder
(`snd1_*`, `sound_*`) are copies from the Ana pack below.

## Ana (`audio/voices/ana/`)

The "Ana" pack is synthetic speech generated with Microsoft Azure AI Speech
(voice `en-US-AnaNeural`) through the author's own Azure resource, under terms
that license the generated audio for distribution with the author's app. It is
shipped as part of this app under the same PolyForm Noncommercial License as
the code. It is not a recording of a real person, and it is not licensed for
use outside this app.

## Owl (`audio/voices/owl/`)

The backup pack (folder `owl`) is synthetic speech generated offline with
[Kokoro](https://github.com/hexgrad/kokoro) (Apache License 2.0). The audio is
shipped under the PolyForm Noncommercial License with the rest of the app.

## Development clips

Clips directly under `audio/` (not in a `voices/` subfolder) are development
artifacts that are not licensed for distribution. They are excluded from the
repository and from every release build.