It's using the same model/engine. I don't have knowledge of the internals, but a different subsystem/set of dedicated resources though for API traffic versus fi

It's using the same model/engine. I don't have knowledge of the internals, but a different subsystem/set of dedicated resources though for API traffic versus first-party apps. One thing to note is there is no separate TTS-phase here, it's happening internally within GPT-4o, in the Realtime API and Advanced Voice.