New realtime voices documented but not available

[f10w]

Hello,

The documentation says that the following three new voices are available in the Realtime API: fable , onyx , nova. However, none of them worked for me.

Has anybody managed to use them?

Thanks a lot in advance for your replies!

[Daysofwonder]

Well, at least on the interactive demo for developers

OpenAI.fm

You can hear them and get an impression of what they sound like. But I haven’t tried using the real-time API yet.

[developer4]

same issue. some of the voices result in error messages

[_j]

The playground has many times showed what actually works, while the documentation is error-filled.

Here’s the realtime playground getting its audio samples in the browser, essentially a list of what is supported:

(along with blocked session tracking and recording in red)