# Notices for the "Your voice" page

This page runs Kyutai's Pocket TTS in the browser. It is made of the parts below; each
stays under its own licence.

## The speech model: Pocket TTS, by Kyutai

- Weights: CC BY 4.0. Source: https://huggingface.co/kyutai/pocket-tts (English, the
  April 2026 model). Paper: https://arxiv.org/abs/2509.06926
- The files under `models/` are derived from those weights: PocketTTS-RAVEN rewrote the
  graphs (delta caches, merged flow) and quantised them; this project compressed them
  with gzip and changed nothing else. Raw file hashes (sha256), as in PocketTTS-RAVEN's
  `webdemo/models/onnx-manifest.json`:
  - `flow_lm_main_delta_flow_int8.onnx` 8300c745f41b611640b542e4ef3dd2203e2c4bc28e32975876ce4a923bd9f1d5
  - `mimi_decoder_delta_int8.onnx` cf6e85b2922db5323515ee8ace74dfa4c18a78565540f7ae37d2c66000094192
  - `mimi_encoder.onnx` 853e2ca623b8782d94c3745ec6133bfdff7ce33d9b11128bd29ea03f28d76e3d
  - `text_conditioner.onnx` 4ecee995fb69f85c7a7493d11f7b5ee15d9950facc7ab3f5c9c49ef1e03847bb
- Kyutai's conditions of use, which apply to this page and to anyone using it:

  > Use of our model must comply with all applicable laws and regulations and must not
  > result in, involve, or facilitate any illegal, harmful, deceptive, fraudulent, or
  > unauthorized activity. Prohibited uses include, without limitation, voice
  > impersonation or cloning without explicit and lawful consent; misinformation,
  > disinformation, or deception (including fake news, fraudulent calls, or presenting
  > generated content as genuine recordings of real people or events); and the generation
  > of unlawful, harmful, libelous, abusive, harassing, discriminatory, hateful, or
  > privacy-invasive content.

  The page asks for the reader's own voice only, and the first sentence read aloud is the
  consent.

## The default voices: Jane and Michael

- `presets/jane.*` and `presets/michael.*` are synthetic voices: voice embeddings made from
  two sample recordings, not the recordings themselves. They stay under CC BY 4.0
  (https://creativecommons.org/licenses/by/4.0/) and are not under the repository's own terms.
- Source: `vctk/p339_023_enhanced.wav` (Jane) and `vctk/p360_023_enhanced.wav` (Michael) in
  Kyutai's TTS voices, https://huggingface.co/kyutai/tts-voices at commit
  323332d33f997de8394f24a193e1a76df720e01a, where they are the samples of Pocket TTS's own
  preset voices of the same names. Kyutai's copies are cleaned versions of the originals
  ("Created using ai-coustics").
- The originals are speakers p339 and p360 of the CSTR VCTK Corpus, version 0.92, by Junichi
  Yamagishi, Christophe Veaux and Kirsten MacDonald, University of Edinburgh, The Centre for
  Speech Technology Research, 2019, https://doi.org/10.7488/ds/2645, CC BY 4.0.
- Changes made here: each recording was trimmed to its speech, brought to this page's speech
  level, resampled to 24 kHz and passed through Pocket TTS's voice encoder in the browser
  (the steps a visitor's own recording goes through); only the resulting embedding is kept.
- The names are Kyutai's labels for its presets. Neither the corpus's speakers, the
  University of Edinburgh nor Kyutai endorses yapbook, and what these voices read is
  generated speech, never a recording of a person.

## The engine: PocketTTS-RAVEN

- MIT. Copyright (c) 2026 VolgaGerm (PocketTTS.cpp), Copyright (c) 2026 Pantelis Kalogiros
  (PocketTTS-RAVEN additions). Source: https://github.com/pkalogiros/pocket-tts-raven at
  commit 983c9095b5ef6dbcb7cd5c57301ffee78633c4df. Licence text: `RAVEN-LICENSE.txt`.
- `src/engine/*.js` are unchanged copies of that project's `webdemo/src/engine/`.
- `src/voice-worker.js` is adapted from its `webdemo/src/tts-worker.js` (compressed models
  kept in the browser's cache; no voice encoder in the reading Worker).
- `models/spm_vocab.json` and `models/bos_before_voice.npy` come from the same project
  (model-derived, CC BY 4.0).

## The runtime: ONNX Runtime Web

- MIT. Copyright (c) Microsoft Corporation. Version 1.27.0, loaded from
  https://cdn.jsdelivr.net/npm/onnxruntime-web@1.27.0/dist/ (see `src/runtime.js`).

## Everything else in this folder

Written for yapbook and under the repository's own terms.