I really don’t understand this. But why real-time voices on ChatGPT and API are totally different? I’m not talking about characters. Their talking abilities are completely different. The voice on the APi does not sound like a real person’s voice. It can’t arrange its tone for questions or similar. The difference is like the difference between GPT 3 and GPT 4. it’s huge!
I don’t understand why they don’t just release the same thing that we use on the Chat GPT App.
Agree. Why is nobody talking about this?
Hi there , totaly agree, we need better voice for the real time API , we build assistant IA for inbound and outbound call, all the clients said to us ‘‘the voice are not good considering the competitors’’
If some of you have anwser or any tips , i take it !
thanks
Just tried to imbed realtime into my app and it is shockingly poor. Not sure why any team would judge this product as being production ready. Very limited, clunky, and the voice options are shallow. Spent hours trying to get it to be remotely credible and have given up. Open AI should withdraw this product until they get a genuine solution reday.
Thanks for the feedback on naturalness and intonation. The Realtime API has voice-delivery controls you can try: https://developers.openai.com/api/docs/guides/realtime-models-prompting. We haven’t confirmed parity with ChatGPT’s voice experience. We’ll pass along the request; no timeline to share.