VoiceCut public Studio demo audio — provenance and attribution
Generated: 2026-08-28

Purpose
-------
These are purpose-made synthetic recordings for the public VoiceCut Studio
fixture. The fictional English and Ukrainian scripts were written specifically
for this demo. They contain no customer or user recording, personal data,
confidential conversation, commercial secret, music, or third-party programme
material.

The fixture illustrates how VoiceCut connects playback, speakers, transcript
timecodes, CHECK markers, and proposed corrections. It is not audio uploaded to
VoiceCut and it is not the output of the live VoiceCut transcription backend.

Generation record
-----------------
Generator: Piper 1.4.1, run locally
Repository source script: apps/web/scripts/generate-demo-audio.py
Machine-readable manifest: /demo/audio-manifest.json

Files:
- /demo/voicecut-studio-demo-en.wav
  WAV PCM signed 16-bit little-endian, 22050 Hz, mono
  Duration: 54.599 seconds
  Size: 2,407,856 bytes
  SHA-256: 4734d7b88d4a883c94c2ef2e2d6edc0eef40fa29f1b477ea3ea0c6ed7ed78b78
- /demo/voicecut-studio-demo-uk.wav
  WAV PCM signed 16-bit little-endian, 22050 Hz, mono
  Duration: 74.338 seconds
  Size: 3,278,368 bytes
  SHA-256: 0a4e38a773bfcba01efad1e06b8c2f3ab0d692e3982edb03aade655224daca86

Voice models and published licence records
-------------------------------------------
Review date: 2026-08-28
The rhasspy/piper-voices repository is marked MIT:
https://huggingface.co/rhasspy/piper-voices
The individual model cards below identify the datasets used to train each
selected voice and their dataset licences. This record states those published
facts; it is not a broader legal warranty about generated audio.

English
  Piper model: en_US-libritts_r-medium
  Selected readers:
  - Piper speaker slot 0 / LibriTTS-R reader 3922: Ashley Candland
  - Piper speaker slot 2 / LibriTTS-R reader 4535: Brett W. Downey
  - Piper speaker slot 3 / LibriTTS-R reader 6701: Dayle
  Model card:
  https://huggingface.co/rhasspy/piper-voices/blob/main/en/en_US/libritts_r/medium/MODEL_CARD
  Dataset: LibriTTS-R, OpenSLR SLR141
  https://www.openslr.org/141/
  Dataset licence: Creative Commons Attribution 4.0 International (CC BY 4.0)
  https://creativecommons.org/licenses/by/4.0/
  Use in this fixture: the model synthesized new purpose-written text with
  adjusted speech length; no OpenSLR source recording is bundled here.
  Dataset citation requested by OpenSLR:
  Yuma Koizumi, Heiga Zen, Shigeki Karita, Yifan Ding, Kohei Yatabe,
  Nobuyuki Morioka, Michiel Bacchiani, Yu Zhang, Wei Han, and Ankur Bapna,
  "LibriTTS-R: A Restored Multi-Speaker Text-to-Speech Corpus," 2023.

Ukrainian
  Piper models:
  - uk_UA-tetiana-high
    https://huggingface.co/rhasspy/piper-voices/blob/main/uk/uk_UA/tetiana/high/MODEL_CARD
  - uk_UA-mykyta-high
    https://huggingface.co/rhasspy/piper-voices/blob/main/uk/uk_UA/mykyta/high/MODEL_CARD
  - uk_UA-oleksa-high
    https://huggingface.co/rhasspy/piper-voices/blob/main/uk/uk_UA/oleksa/high/MODEL_CARD
  Dataset named by the model cards: Ukrainian TTS Datasets
  https://github.com/egorsmkv/ukrainian-tts-datasets
  Dataset licence named by each model card: Apache License 2.0
  https://www.apache.org/licenses/LICENSE-2.0

Integrity
---------
The manifest records the text spoken by every segment, the selected model and
speaker, and exact start/end times. Repository tests bind that manifest to the
WAV hashes, byte lengths, format, durations, localized transcript, and CHECK
suggestions. ffprobe confirmed the format and durations; ffmpeg decoded both
complete files with zero errors at the 2026-08-28 source checkpoint.