Voice in your app

Voice for your mobile app.

Your app gets a voice: a tutor that talks a student through a problem, a guide that answers in the user’s own language, a coach the user can interrupt. Your server starts each call with one request and passes the app a join token; the app connects with LiveKit’s SDK and shows live captions.

On a call.

What the caller hears and what happens behind it, step by step.

  1. The user taps to talk. Your app asks your server for a call.
  2. Your server starts one with its API key and the user’s details as variables.
  3. The app joins with the token it is given and turns on the microphone.
  4. The agent greets the user by name; captions and its listening or speaking state drive your interface.
  5. When the user hangs up, your server gets the transcript.

What you set up.

All of it in the studio, or through the API.

  • Your server. One request to start a call. Your API key never leaves it.
  • The app. LiveKit’s SDK for iOS, Android, React Native or the web: join, send the microphone, play the agent.
  • Variables. The user’s name, level or history, so every call starts already knowing them.

A brief to start from.

Paste it into the agent’s instructions and make it yours. The business in it is an example.

You are Mira, a patient English-speaking partner in the Fluent app. The learner’s name is {{name}} and their level is {{level}}. Talk about whatever they choose, keep your sentences short, and gently repeat a corrected version when they make a mistake.

Build it.

The guide and the worked example, with all of the code.

Questions.

Which platforms?

Anywhere LiveKit has a client SDK: the web, iOS, Android, React Native and more. Calls run on LiveKit.

Can the user interrupt it?

Yes. When the user starts talking the agent stops mid-word and listens, and a short “okay” is not taken as an interruption.

Other things to build.

Give your agent a voice today.

Build one in the studio and talk to it in your browser. Sign-up credit is there for your first test calls.