Every voice agent has one voice and one language. The voice picker lets you preview every option with your own greeting before saving, so you can hear exactly what your callers will hear. This article walks through the voice families and the full list of supported languages.
The 30 voices
Voices are powered by Gemini and grouped into seven families. Each voice can speak every supported language — picking a voice does not restrict your language choices.
| Family | Best for | Voices in this family |
|---|---|---|
| Authoritative | B2B receptionists, financial services, healthcare, legal | Charon, Rasalgethi, Sadaltager, Kore, Orus, Alnilam |
| Bright/Upbeat | Consumer brands, retail support, sales | Zephyr, Puck, Autonoe, Laomedeia, Sadachbia |
| Energetic | Younger audiences, entertainment, high-volume support | Fenrir, Pulcherrima, Leda |
| Casual | Friendly consumer support, hospitality, food & beverage | Aoede, Callirrhoe, Umbriel, Zubenelgenubi, Achird |
| Soft/Warm | Wellness, after-hours support, premium service brands | Achernar, Vindemiatrix, Sulafat, Enceladus |
| Clear/Smooth | Technical support, IVR-replacement, formal corporate | Iapetus, Erinome, Algieba, Despina, Schedar |
| Distinctive | Memorable, character-driven brands | Algenib, Gacrux |
The default voice is Kore (Authoritative, female) — a safe pick if you are unsure.
How to preview a voice
On the Voice & language tab of a voice agent, click any voice card to play a sample. The sample uses your agent's actual begin message in the chosen language, so you hear exactly what callers will hear on the first second of a call. Previewing does not cost anything.
Match the brand tone, not the country. A premium watch brand does not change voice between English and German — it changes voice between premium and friendly. Pick the family that fits your brand and use the same voice across every language agent you set up.
The 17 languages
| Language | Code |
|---|---|
| English (US) | en-US |
| English (UK) | en-GB |
| English (Australia) | en-AU |
| English (Ireland) | en-IE |
| Hungarian | hu-HU |
| German | de-DE |
| French | fr-FR |
| Spanish (Spain) | es-ES |
| Spanish (Latin America) | es-MX |
| Portuguese (Brazil) | pt-BR |
| Portuguese (Portugal) | pt-PT |
| Italian | it-IT |
| Dutch | nl-NL |
| Polish | pl-PL |
| Czech | cs-CZ |
| Slovak | sk-SK |
| Romanian | ro-RO |
What happens if a caller switches language mid-call
A voice agent stays in its configured language for the whole call. If a caller switches to another language partway through:
- The agent replies in its configured language, but may slow down or simplify wording.
- Brief code-switching — for example, a Spanish caller dropping in an English brand name — is handled transparently with no extra configuration.
- For full multilingual support, create one voice agent per language and give each agent its own phone number. Callers self-select by dialing the right number.
Each agent's greeting is pre-generated for instant pickup. If you ask a single agent to switch language mid-call, there is a noticeable beat of silence while the model reorients — one agent per language avoids this entirely.
Transcript fallback for non-English
Transcription uses Gemini Live's native speech-to-text for the configured language. For some non-English calls where Gemini transcription is weaker, Yaplet automatically falls back to Deepgram for the transcript only — the spoken side of the conversation is unaffected. This is automatic, with no configuration needed.
What to read next
Once you have picked a voice and language, the next step is making sure the agent has good knowledge to draw from. See Connect knowledge to your voice agent.