Little Meadow neural voice credits

The voices in this game are synthesized locally by Kokoro. They are not human
recordings and do not represent or endorse any real person.

Kokoro-82M: hexgrad / rzvzn — Apache License 2.0.
https://huggingface.co/hexgrad/Kokoro-82M
ONNX conversion and voice embeddings: ONNX Community / Xenova — Apache 2.0.
https://huggingface.co/onnx-community/Kokoro-82M-v1.0-ONNX
Pinned revision: 1939ad2a8e416c0acfeecc08a694d14ef25f2231
Portuguese: pf_dora (lessons), pm_alex and pm_santa (comparison previews).
English: af_heart, the voice named Buddy in DrawHop.

Kokoro's upstream training-data credits are preserved here:
Koniwa — https://github.com/koniwa/koniwa — CC BY 3.0, per the model card.
SIWIS — https://datashare.ed.ac.uk/handle/10283/2353 — CC BY 4.0.
Architecture: StyleTTS2 by Yinghao Aaron Li et al. (MIT code), and iSTFTNet.
https://github.com/yl4579/StyleTTS2

The game distributes generated WAV clips, not the inference model or authoring
libraries. The separately installed authoring tools use ONNX Runtime (MIT),
Misaki (Apache 2.0), spaCy (MIT), and eSpeak-NG / phonemizer (GPL) for phoneme
conversion. Do not bundle the authoring virtual environment into the game;
redistributing these tools carries their separate license obligations.

No model training, voice cloning, remote inference, or runtime downloads were
used to make these clips. Changes: Little Meadow text, local synthesis,
conservative outer-silence trimming and peak limiting. Full model license:
LICENSE-Kokoro.txt. Build provenance: voice-info.json.