Audio Customization

How the agent sounds is controlled by the text-to-speech plugin in the pipeline. Each TTS provider exposes a voice plus tuning knobs; set them on the plugin’s constructor.

Choose a voice

Pass the provider’s voice identifier. Voices and the exact option names vary by provider. See each TTS plugin page for the full list.

from zrt.plugins import CartesiaTTS

tts = CartesiaTTS(
    voice="<voice-id>",
    speed=1.0,
)

# Pipeline(tts=tts, ...)

Common tuning knobs

The exact set depends on the provider, but most TTS plugins support some of:

Knob	Effect
`voice` / `speaker`	The voice identity.
`speed` / `pace`	How fast the agent talks.
`pitch`	Higher or lower voice.
`emotion`	Emotional tone, where the provider supports it.
`language`	Spoken language / accent.
`sample_rate`	Output audio sample rate.

For voice agents, lower latency usually beats richer prosody. Prefer a fast model and streaming synthesis, and keep replies short.

Speaking on demand

Make the agent speak a specific line at any time with session.say():

await self.session.say("Let me look that up for you.")

To fix how specific words are pronounced (acronyms, brand names), transform the text before it reaches TTS. See Text Transformation.

References

Python

Examples

Enhanced Pronunciation

Customize pronunciation of names and terms.

Multilingual Demo

Switch speech languages within an agent.

SDK Reference

Speech Generation

Speech Generation in the Python API reference.

TTS

TTS in the Python API reference.

Overview Background Audio

Get started

Modalities

Configure Pipeline

Tools and Capabilities

Turn Detection and Interruptions

Telephony Agents

Context Management

Observability and Monitoring

Choose a voice

Common tuning knobs

Speaking on demand

References

Examples

Enhanced Pronunciation

Multilingual Demo

SDK Reference

Speech Generation

TTS

​Choose a voice

​Common tuning knobs

​Speaking on demand

​References

​Examples

Enhanced Pronunciation

Multilingual Demo

​SDK Reference

Speech Generation

TTS

Choose a voice

Common tuning knobs

Speaking on demand

References

Examples

SDK Reference