LotM: Beyond the Veil — original madness voice assets, 0.28.0

44 locally generated voice takes, two for each of the 22 Pathways.
66 original on-screen lines. These are game adaptations, not quotations from the novels.
Fool explicitly uses the user-requested Hornacis and Flegrea motifs.

Speech model: Kokoro-82M v1.0 by hexgrad, Apache-2.0 weights.
Official model and training-attribution record: https://huggingface.co/hexgrad/Kokoro-82M
Inference/export tool: kokoro-onnx 0.6.0, MIT, https://github.com/thewh1teagle/kokoro-onnx
Exported model and voice vectors: https://github.com/thewh1teagle/kokoro-onnx/releases/tag/model-files-v1.1
Default model voices: af_nicole and am_echo. No requested impersonation or custom voice clone.
Project processing: phonetic-envelope breath texture, band shaping, short diffused echoes,
soft crest limiting, conservative gain, start/end fades, 24 kHz mono Vorbis encoding.
Hashes and measured levels: madness-manifest.json.
Reproducible authoring script: scripts/export_madness_audio_0280.py in the source project.

Only finished Ogg sound assets are included in the game. No model weights, Python runtime,
online synthesis service, recording capability, or network speech dependency is shipped.
Source WAV masters and the synthesis installation are kept in the isolated production workspace.
