Skip to main content
Use Fish Audio for text-to-speech in your LiveKit agent and Anam to turn that speech into an interactive avatar. You choose the speech-to-text (STT) and large language model (LLM); LiveKit runs the conversation and delivers the avatar’s audio and video to your frontend.

How it works

Fish Audio turns the LLM’s replies into speech. The Anam LiveKit plugin sends that audio to Anam, which joins the LiveKit room as a participant and publishes synchronized audio and avatar video. Your frontend renders the avatar participant’s tracks.
Fish Audio runs inside your LiveKit agent. You don’t import a voice into Anam for this Fish Audio LiveKit implementation.

Choose how to connect Fish Audio

You can choose between two ways to implement Fish Audio with Anam. LiveKit Inference is the quickest: LiveKit serves Fish Audio’s voices, so you only need LiveKit and Anam credentials and LiveKit bills the usage. The Fish Audio plugin connects to Fish Audio with your own API key, so you’re billed by Fish Audio directly and can use any voice on your account. The Quickstart uses LiveKit Inference. Configuration shows how to switch to the plugin.

See it in action

Try the Fish Audio integration demo to talk to an Anam avatar that speaks with Fish Audio’s Drama 3 voice model, powered by LiveKit.

Resources

Quickstart

Build a Python agent and connect a React frontend

Configuration

Voices, models, Fish Audio options, and deployment

LiveKit Fish Audio docs

Fish Audio models, voices, and parameters in LiveKit

Anam LiveKit plugin

Avatar options, Director Notes, and advanced examples
Last modified on November 17, 2026