# Build a voice agent that speaks Yoruba, Hausa and Pidgin

URL: https://bridge.wekoya.tech/guides/voice-agent-yoruba-hausa-pidgin
Updated: 2026-11-08

**In short:** Use Gemini Live through Bridge by Wekoya: one WebSocket where the model hears the caller and answers by voice in English, Yoruba, Hausa or Pidgin, with its first sound about 1.3 seconds after the caller stops. It can call your tools mid-conversation and is billed in naira by the minute heard and spoken.

## How it fits together

1. Your server asks Bridge for a short-lived ticket, so the browser never holds your key.
2. The browser opens the WebSocket with the ticket and streams the microphone.
3. The model answers by voice; when it needs your data (an order, a booking) it calls one of your tools and your code answers.
4. After each turn Bridge reports the cost so far, in naira.

The [live voice guide](https://bridge.wekoya.tech/docs/realtime) has the full server and browser code, tested end to end, and the rules that keep a call smooth. Prices per minute are on the [models page](https://bridge.wekoya.tech/models).

## When to build it differently

If you need a particular voice or a chat model of your own choosing, chain three steps instead: live transcription, a fast chat model, and text to speech. See [voice agents](https://bridge.wekoya.tech/docs/voice-agents). Speech to text with nova-3 does not support Yoruba, Hausa or Igbo; Gemini Live does understand Yoruba, Hausa and Pidgin.

[Create a free account](https://bridge.wekoya.tech/sign-up)

## Common questions

**Which Nigerian languages does the voice agent understand?**

Gemini Live on Bridge understands and answers in English, Yoruba, Hausa and Pidgin.

**How fast does it answer?**

With gemini-3.1-flash-live-preview, the first sound comes about 1.3 seconds after the caller stops speaking.

## Read next

- [Live voice docs](https://bridge.wekoya.tech/docs/realtime)