Interface SpeechSteeringOptions

Steers verbal delivery and non-verbal sounds in generated speech.

Every key is a sparse override on the default (full sound vocabulary, light fillers): the expressive instructions already tell the LLM to match its delivery to the register of the moment, so most agents need no steering at all — set a key only to take an option away regardless of context.

interface SpeechSteeringOptions {
    disfluencies?: boolean;
    nonverbalSounds?: boolean | NonverbalOptions;
    pace?: "fast" | "slow" | "normal";
}

Properties

disfluencies? nonverbalSounds? pace?

Properties

Optional disfluencies

disfluencies?: boolean

Filler words such as "um" / "uh". On by default (DEFAULT_SPEECH_STEERING_OPTIONS); set false to opt out.

Optional nonverbalSounds

nonverbalSounds?: boolean | NonverbalOptions

Which non-verbal sounds the TTS may make. true (and omitting the key) keeps the provider's full vocabulary, false disables every sound, and a NonverbalOptions object toggles per category (omitted categories stay enabled).

Optional pace

pace?: "fast" | "slow" | "normal"

Settings

Member Visibility

Theme

On This Page

disfluenciesnonverbalSoundspace