Build your agent
Speaking
Text-to-speech: the voices your agent can speak with, and how to choose one.
| Provider | service | Best for |
|---|---|---|
| Deepgram Aura | deepgram | Low-latency conversational voices |
| Sarvam | sarvam | Natural Indian voices |
| ElevenLabs | elevenlabs | Expressive, cloneable voices |
google | Chirp 3 HD voices in many languages | |
| Resemble | resembleai | Custom cloned voices |
Getting it right
For live calls pick a low-latency model (Deepgram Aura, Sarvam Bulbul, ElevenLabs Flash). Preview voices on Voices before you choose, and try speed between 1.0 and 1.1: slightly brisk sounds more natural on the phone.
Next steps
- Deepgram Aura: low-latency conversational voices.
- Sarvam: natural indian voices.
- ElevenLabs: expressive, cloneable voices.
- Google: chirp 3 hd voices in many languages.
- Resemble: custom cloned voices.