Skip to main content
OpenAI GPT Live is a speech-to-speech model that listens and speaks in one step, enabling faster and more natural voice assistant conversations. Unlike traditional text-to-speech systems, GPT Live processes audio input directly and generates spoken responses without an intermediate text transcription step.

​
Available Voices

GPT Live offers 22 voices:
  • Alloy
  • Ash
  • Ballad
  • Breeze
  • Cedar
  • Coral
  • Cove
  • Echo
  • Ember
  • Fable
  • Glimmer
  • Juniper
  • Marin
  • Meadow
  • Nova
  • Reef
  • Ridge
  • Sage
  • Shimmer
  • Vale
  • Verse
  • Whisper
You can preview any voice in the Voice library before switching.

​
How to Configure GPT Live

1

Open Voice Settings

Open your assistant and navigate to Voice settings, then Voice library.
2

Select Provider and Voice

Set Provider to OpenAI GPT Live and choose a voice from the catalog.
3

Choose Chat & Tools Model

In the assistant header, select a separate AI model to handle tool actions (bookings, lookups, transfers) and chat replies. This ensures your assistant retains all its capabilities while using the GPT Live voice.
4

Publish

Click Publish to make your changes live.

​
Important Considerations

  • Switching to or from a GPT Live voice requires a full Publish operation.
  • GPT Live voices cannot be used together with a connected Flow.
  • Transcription and voice-tuning settings do not apply when using GPT Live, since the model handles both hearing and speaking directly.

​
Chat & Tools Model

Because GPT Live is a speech-to-speech model optimized for conversation, you must select a separate Chat & tools model to handle:
  • Tool function calls (appointment bookings, contact lookups, call transfers)
  • Chat message replies
  • Structured data extraction
This separation ensures your assistant maintains full functionality while benefiting from GPT Live’s natural conversational flow.