On October 30, we introduce new voices on chat completions to replace others. These voices are now included in the changelog and are integrated into the Python library.
Python Commit Details
The new voices are enforced and allowed only by Python openai-1.53.0. The gpt-4o-audio-preview, also known as gpt-4o-audio-preview-2024-10-01 model, does not currently support these new voices. It is the only model listed by the models endpoint or in project permissions or limits (with a new key generated).
Despite this, the API is rejecting the new voices and still listing the old ones. If you are not using OpenAI’s library, you have the flexibility to send whatever voices you prefer. However, functionalities like fable, etc., continue to work seamlessly.
Realtime API Playground
There is a preview of the new voices and the omission of the others in the realtime API playground. The new voices are showcased with over-the-top enthusiasm and are working flawlessly, infinitely generating dialogues like "¡Hola! ¿Qué tal? ¿En qué".
This platform is powered by Discourse and is best viewed with JavaScript enabled.
