Skip to content
sensor-humor

Voice Integration

sensor-humor pairs with mcp-voice-soundboard to deliver spoken comedy with mood-appropriate prosody.

How it works

Section titled “How it works”

Each mood maps to a Piper TTS voice and 4 prosody knobs:

Knob What it controls Effect on comedy
length_scale Speech speed (>1 = slower) Dry/cynic drag; zoomer rushes
noise_scale Expressiveness (0 = flat, 1 = animated) Dry is monotone; chaotic is erratic
noise_w_scale Phoneme timing variance (0 = metronomic, 1 = erratic) Controls whether words land at even intervals
volume Loudness Cynic is quiet; zoomer is loud

Mood-to-voice mapping

Section titled “Mood-to-voice mapping”
Mood Piper Voice length noise noise_w vol
dry en_GB-alan-medium 1.15 0.3 0.3 0.9
roast en_US-ryan-high 0.95 0.667 0.8 1.0
chaotic en_US-lessac-high 0.88 0.8 0.9 1.1
cheeky en_GB-cori-high 1.05 0.5 0.6 0.95
cynic en_GB-alan-medium 1.25 0.2 0.2 0.8
zoomer en_US-lessac-high 0.90 0.85 0.9 1.15

Setup

Section titled “Setup”
  1. Install Piper voice models:
Terminal window
# Download to your model directory
# Required: alan-medium, ryan-high, lessac-high, cori-high
  1. Start voice-soundboard with Piper:
Terminal window
VOICE_SOUNDBOARD_ENGINE=piper \
VOICE_SOUNDBOARD_PIPER_MODEL_DIR=/path/to/piper/models \
npm start
  1. After any sensor-humor tool call, speak the result:
voice_speak({ text: result.roast, mood: "roast" })

Kokoro fallback

Section titled “Kokoro fallback”

If VOICE_SOUNDBOARD_ENGINE is not set, the default Kokoro backend is used. Kokoro supports voice selection and speed control but not full prosody (noise_scale, noise_w_scale, volume are ignored). The comedy still works — just without the prosodic separation between moods.

Why Piper over Kokoro

Section titled “Why Piper over Kokoro”

Kokoro strips SSML and only respects speed + voice selection. Piper’s ONNX inference exposes all 4 prosody knobs natively via SynthesisConfig, giving real vocal differentiation:

  • Dry and cynic use the same voice (alan) but cynic is measurably slower, flatter, and quieter
  • Chaotic has high timing variance — words don’t land at even intervals, which reads as erratic
  • Roast has natural expressiveness that gives verdict labels a confident lift
  • Zoomer is fast and loud — streamer energy with brisk pacing