> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sentivue.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Gemini TTS

> Connect Gemini TTS to SentiVue Talk

Gemini TTS provides the **voice** that speaks your agent's responses. It takes a style prompt — plain-language direction for how lines should be delivered.

## What you need

| Field                | Required |
| -------------------- | -------- |
| Service-account JSON | Yes      |

## Getting a service-account JSON

1. In the [Google Cloud Console](https://console.cloud.google.com/), select or create a project.
2. Enable the **Cloud Text-to-Speech** API on it.
3. Go to **IAM & Admin → Service Accounts** and create one.
4. Open its **Keys** tab, choose **Add Key → Create new key**, and select **JSON**.
5. The file downloads once. Store it securely and treat it as a secret.

Upload the whole file, including `private_key` and `client_email`.

## Adding the connection

1. Go to **Connections** and choose **Gemini TTS**.
2. Use **Upload JSON** to attach the service-account file.
3. Save.

## What you can select afterwards

Choose your model on [Model Configurations](/talk/configurations/inference-providers).

**Model** — use a `flash` model on calls. `pro` is slower and the delay is audible to the caller. See Google's [speech generation models](https://ai.google.dev/gemini-api/docs/speech-generation).

**Voice** — Kore is the default. Google's [voice list](https://ai.google.dev/gemini-api/docs/speech-generation#voices) describes how each one sounds.

**Language** — see Google's [speech generation models](https://ai.google.dev/gemini-api/docs/speech-generation) for current coverage.

**Style prompt** — optional. Describes delivery, for example `Speak warmly and unhurriedly, like a receptionist who is not in a rush`. Shapes tone, pace, and accent for the whole call. Blank gives the default delivery.

**Region** — optional server region, such as `us-central1`. Set per workflow, not on the connection.

## Which Google voice connection to use

| You want                                                                               | Use                                                              |
| -------------------------------------------------------------------------------------- | ---------------------------------------------------------------- |
| Style-directed delivery                                                                | **Gemini TTS**                                                   |
| Chirp 3 HD voices, widest language coverage, voice and transcription on one credential | [Google Cloud](/talk/configurations/providers/google-cloud)      |
| One model that hears and speaks, replacing the transcriber                             | [Gemini Live](/talk/configurations/providers/google-gemini-live) |

Each is a separate connection with its own credential.

## Troubleshooting

<AccordionGroup>
  <Accordion title="I have an AI Studio API key, not a service account">
    Gemini TTS does not accept one. An AI Studio key drives the language model and belongs on the [Gemini](/talk/configurations/providers/google-gemini) connection.
  </Accordion>

  <Accordion title="Voice fails at call time although the JSON was accepted">
    Enable the **Cloud Text-to-Speech** API on the service account's project. Without it the credential saves and the call fails.
  </Accordion>

  <Accordion title="The service-account key will not upload">
    It must be the JSON file, not P12, and the whole file rather than one field from it.
  </Accordion>
</AccordionGroup>