# Audio v12 credits and provenance

The unchanged paper, ceramic, glass, eating, monster, metal and pouring cues remain in `../audio-v11/`; their existing credit file applies.

## Source sound effects (CC0 1.0)

- `lightSwitch.mp3`: **Light Switch Turn On And Off SFX**, FunnyDude — https://opengameart.org/content/light-switch-turn-on-and-off-sfx-0 . Mechanical off-switch take, trimmed. The source does not identify the mechanism as a pull cord; this file is not represented as a verified pull-cord field recording.
- `cough.mp3`: **Cough**, poorenglishjuggler — https://freesound.org/people/poorenglishjuggler/sounds/155650/ . Original source excerpt 0.185–1.240 seconds; only two coughs, no third attack.
- `squish.mp3`: **Squish Sounds Effects**, EZduzziteh — https://opengameart.org/content/squish-sounds-effects . `squish_06.mp3` excerpt. This is a dedicated bug-squashing source asset; its recording method is unspecified by the author.
- `curtain.mp3`: **Fabric Rustling**, Iochi Glaucus — https://opengameart.org/content/fabric-rustling . Real fabric recording; two passages crossfaded, low rumble removed, no rail knocks or impacts.
- `jelly.mp3`: **Bubble Sound Effects**, Brian MacIntosh / BMacZero — https://opengameart.org/content/bubble-sound-effects , plus a quiet **Squish Sounds Effects** texture by EZduzziteh above. Edited soft bubble body with a quiet rebound; no spring-doorstop sample.
- `jellyPop.mp3`: **Squish Sounds Effects**, EZduzziteh, plus **6 Short water splashes**, ezwa (submitted by qubodup) — https://opengameart.org/content/6-short-water-splashes . Wet pop with a short splash tail; separate from the bounce cue.

CC0: https://creativecommons.org/publicdomain/zero/1.0/ . Attribution is provided voluntarily. Individual source URLs, edits, durations and final file hashes are in `manifest.json`.

## Generated dialogue — not live voice recordings

The six `guestMale*.mp3` / `guestFemale*.mp3` clips were generated locally for this website with **Kokoro-82M v1.0**, using its standard British English **bm_george** and **bf_emma** voice embeddings. These are distinct voice banks at normal playback pitch, not one voice artificially shifted. No named person's voice was cloned for this project.

Each character says exactly **Hey!**, **Shh!**, or **Go away!**. The shush was synthesized from the /ʃː/ phoneme rather than spelling the letters. All generation uses speed 1 and normal playbackRate 1; only EQ, fades and level mastering were applied afterward.

Model: https://huggingface.co/hexgrad/Kokoro-82M

Voice documentation: https://huggingface.co/hexgrad/Kokoro-82M/blob/main/VOICES.md

Model weights: Apache License 2.0 — https://www.apache.org/licenses/LICENSE-2.0

ONNX engine: https://github.com/thewh1teagle/kokoro-onnx (MIT). The model/engine licenses describe the generation tools; no model weights or engine are shipped with this website. Export release and exact hashes are recorded in `manifest.json`.

## Validation scope

All twelve final MP3s were decoded and checked for finite samples, non-silence, duration and safe digital peak. Metadata/phoneme inputs verify the intended wording. Subjective listening review was not available in this executor; no claim of having auditioned them is made.