Skip to content

Voices

Soundboard Plugin ships with 12 curated voices covering male, female, American, and British accents. Each voice has a distinct personality suited to different use cases.

VoiceIDGenderStyle
Fenriram_fenrirMPowerful, authoritative (default)
Ericam_ericMEnergetic, urgent
Liamam_liamMWarm, conversational
Onyxam_onyxMDeep, steady
Aoedeaf_aoedeFClear, expressive
Jessicaaf_jessicaFProfessional, neutral
Skyaf_skyFBright, friendly
Alicebf_aliceFBritish, composed
Emmabf_emmaFBritish, warm
Isabellabf_isabellaFBritish, refined
Georgebm_georgeMBritish, formal
Lewisbm_lewisMBritish, measured

Voice ID prefixes: am_ = American male, af_ = American female, bm_ = British male, bf_ = British female.

Presets map common use cases to a voice and speed combination:

PresetVoiceSpeedUse for
defaultam_fenrir1.0xPowerful, commanding (Fenrir)
narratoram_fenrir0.95xCalm, powerful pace (Fenrir)
announceram_eric1.1xConfident, bold (Eric)
storytellerbf_emma0.9xRefined, expressive (Emma)
friendlyam_liam1.0xWarm, approachable (Liam)
professionalaf_jessica1.0xClear, business-ready (Jessica)

The plugin detects 8 emotions from context and adjusts voice delivery:

EmotionDefault voiceSpeedWhen it triggers
neutralbm_george1.0xDefault, no strong signal
seriousbm_george1.0xError reports, warnings
friendlyam_liam1.0xPositive interaction, greetings
professionalaf_jessica1.0xBusiness context, formal tone
calmbm_george0.95xExplanations, walkthroughs
joyam_liam1.06xCelebrating success, positive results
urgentam_eric1.12xTime-sensitive notifications, alerts
whisperam_onyx0.92xQuiet, subtle, ambient speech

Emotion detection is automatic based on the text content and context. You can also pass an explicit emotion parameter to voice.speak to override the detection. Unknown emotion names are downgraded to neutral with a warning.

All configuration is via environment variables. No config files.

VariableDefaultDescription
VOICE_SOUNDBOARD_OUTPUT_ROOT{tempdir}/voice-soundboard/WAV output directory
VOICE_SOUNDBOARD_RATE_COOLDOWN_MS0 (disabled)Per-tool rate limit cooldown in milliseconds
VOICE_SOUNDBOARD_RETENTION_MINUTES240Auto-delete WAVs older than this (minutes)
VOICE_SOUNDBOARD_AMBIENT_ENABLED0Enable inner monologue system (set to 1 to enable)

Only the 12 voices listed above are permitted. Passing an unknown voice ID to any tool results in a structured error with a hint listing valid options. This is a security measure to prevent arbitrary voice model loading.