OpenAI offers versatile multilingual voices balancing low latency and good quality. While voices are optimized for English, they perform well across all supported languages.

Consult OpenAI’s Text-to-Speech documentation for more information and audio samples for available voices.

Models

OpenAI offers two TTS models with different quality and latency characteristics:

Model	Description
tts-1	Default - Standard quality, optimized for speed
tts-1-hd	High-definition quality, improved audio fidelity

Voices

OpenAI provides 6 fully multilingual voices optimized for natural-sounding speech:

Voice	Description
alloy	Neutral, balanced voice
echo	Warm, expressive voice
fable	British accent, articulate
onyx	Deep, authoritative voice
nova	Energetic, friendly voice
shimmer	Soft, gentle voice

Languages

All OpenAI voices are fully multilingual and automatically adapt to the input text language.

Consult OpenAI’s supported languages resource for an up-to-date list.

Usage

OpenAI voice IDs conform to the following format:

openai.<voice>:<model>

Parameters:

• voice (required): One of: alloy, echo, fable, onyx, nova, shimmer
• model (optional): tts-1 (default) or tts-1-hd

Examples:

openai.alloy
openai.nova:tts-1-hd
openai.fable:tts-1

---

Examples

Learn how to use OpenAI voices on the SignalWire platform.

SWML

RELAY Realtime SDK

Call Flow Builder

Use the languages SWML method to set one or more voices for an AI agent.

version: 1.0.0
sections:
  main:
  - ai:
      prompt:
        text: Have an open-ended conversation about flowers.
      languages:
        - name: English
          code: en-US
          voice: openai.alloy

Alternatively, use the say_voice parameter of the play SWML method to select a voice for basic TTS.

version: 1.0.0
sections:
  main:
  - set:
      say_voice: "openai.alloy"
  - play: "say:Greetings. This is the Alloy voice from OpenAI's text-to-speech model."