# Babble voice samples — offline renders of the M5 system

Rendered through the real scheduler + formant phoneme synth (the same
pure code the client runs), mixed offline. Deterministic: these exact
takes are what every client hears for the same line + voice.

## Same line, four derived voices
`voice-alice/bruno/cleo/dax.opus` — "Hello traveler! Care to trade, or
are you just passing through?" Voices derived purely from the identity
hash (no stored data): differing base pitch (130–270 Hz), jitter,
rate, formant width.

## Intonation demo (one voice, cleo)
- `intonation-statement.opus` — "We found the door." (falls)
- `intonation-question.opus` — "We found the door?" (rises)
- `intonation-exclaim.opus` — "We found the door!" (lifted + quicker)

## Parameter extremes (manual overrides, sanitized)
- `voice-tiny-quick.opus` — basePitch 340, rate 11, formant 1.35 — a
  small quick creature.
- `voice-huge-slow.opus` — basePitch 75, rate 3.5, formant 0.72,
  roughness 0.8 — something large.
- `voice-everychar.opus` — cadence 1 (every-char blips, the busiest
  setting; derived player voices pin cadence 0 = syllables-only).

In-game: Settings → Voice Preview plays your own derived voice; emotes
babble stock phrases; entities speak via the server Speak hook.