ZoneVoice English showcase recordings
Created: September 26, 2026

All scripts are original demonstration material. People, projects, dates, and
commitments are fictional. These are not customer recordings or testimonials.
Voices are AI-generated using stock Kokoro voices, not imitations of a person
named in the scripts.

Speech generation:
- Kokoro-82M by hexgrad: https://huggingface.co/hexgrad/Kokoro-82M
- ONNX conversion: https://huggingface.co/onnx-community/Kokoro-82M-v1.0-ONNX
- kokoro-js 1.2.1: https://github.com/hexgrad/kokoro/tree/main/kokoro.js
- Model and package license: Apache License 2.0
  https://www.apache.org/licenses/LICENSE-2.0
- Voices: af_heart (Maya), am_michael (Noah), af_bella (Leah), bm_george (Owen)
- Local CPU generation, q8 model, 24 kHz mono PCM16 WAV.
- A 350 ms silence separates turns. JSON timestamps come from generated audio
  lengths, not forced alignment or a transcription benchmark.

The .txt files are original scripts. The .md files and scenario JSON contain
AUTHORED reference notes, tasks, answers, and example knowledge states.
They are not measured output from ZoneVoice or an actual MCP conversation.

Screenshots under /images/showcase are captures of the real ZoneVoice desktop
preview processing the synthetic audio. Their results are distinct from the
curated reference notes and may differ. Maya and Owen in the handoff transcript
were manually named in the app after generation. No UI was redrawn or composited.

The separate Atlas Memory screenshots use synthetic text fixtures described in
/guides/meeting-knowledge-base, not these Harbor recordings.

AMI and LibriSpeech were researched as future evaluation sources. None of their
recordings or transcripts are included in this package.