Skip to main content

​
Supported Voice Styles

Choose from our collection of voice styles to personalize your digital human’s voice in the Real-time Digital Human API. Each voice style has unique characteristics suited for different use cases and scenarios. Click the play button to preview each voice style.

​
Accent and Dialect Support

You can control accents and dialects by specifying them in your instructions (system prompt). The model will match the accent/dialect you specify while using the selected voice style. Example: Control dialect for consistent persona: 
instructions: `## Language

Response only in argentine spanish.`
For more details on language and dialect control, see the Building Effective Prompts guide.

​
How to Use

After establishing the WebSocket connection, send session configuration using realtime.input_config in the onopen event handler: 
socket.onopen = () => {
  const config = {
    voice: 'echo' // OpenAI voice styles: alloy, ash, ballad, cedar, coral, echo, marin, sage, shimmer, verse
  };
  
  // Send configuration
  socket.send(JSON.stringify({
    type: 'realtime.input_config',
    data: { content: JSON.stringify(config) }
  }));
};

​
Note

This page lists voice styles supported by the Real-time Digital Human API. For voice styles supported by the Video Synthesis API, see the Video Synthesis Voice Styles page.