Gemini TTS is Google's generative text-to-speech, available on the default Edesy runtime as a pipeline voice. It offers 30 HD voices that work across languages, including Odia. Use it when you want Gemini voices with a separate STT and LLM. If you want one model to listen, think and speak, use Gemini Live native audio instead.
At a glance#
| Item | Value |
|---|---|
| Provider ID | gemini (aliases gemini_tts, gemini_genai_tts) |
| Runtime | Default runtime |
| Voices | 30 HD voices, for example Kore, Puck, Charon, Aoede, Sulafat |
| Languages | Multilingual voices; each voice speaks the language of the text |
| Auto-detection | An agent with ttsProvider: "google" and a Gemini TTS model is routed to Gemini TTS |
| Your own key (BYOK) | Supported with a Google AI Studio key |
Gemini TTS or Gemini Live?#
| Gemini TTS | Gemini Live (native audio) | |
|---|---|---|
| What it does | Speaks text from any LLM | Listens, reasons and speaks in one model |
| STT and LLM | You choose them | Not needed |
| Best for | Mixing Gemini voices with another LLM, or full control of each step | Natural turn-taking with the fewest moving parts |
Set it up#
- Open the agent and go to Voice → Text-to-speech.
- Choose Gemini TTS.
- Pick a voice. Preview a few, since each has a different style (bright, warm, firm and so on).
- Save and place a test call.
{
"ttsProvider": "gemini",
"ttsVoice": "Kore"
}
FAQs#
Is Gemini TTS streamed? The default runtime uses Gemini's batch (non-streaming) synthesis. If the fastest possible first word matters most, compare it with a streaming voice such as Cartesia.
Is there an add-on charge? Gemini TTS is a premium voice. See provider costs.
Can I spread load across several Google keys? Yes. A workspace can add a pool of up to 10 Google AI Studio keys. See BYOK and key pools.
Does it speak Odia? Yes. Gemini TTS is one of the options for Odia agents.