ChatGPT’s new voice models can both listen and talk, just like a human
New voice models
What you should particularly notice as a user is that ChatGPT’s voice sounds a bit more human and speaks in a more conversation-like manner. The advantage of a voice that already sounds more human is that you are more inclined to talk to it in a natural way, but something has also been done, for example, in the choice of words, which should make the chatbot more inviting to talk to. In the background, something called ‘continuous interaction’ has been added and that ensures that AI can both listen and talk. That’s a big change because it means you can talk over him and he can both keep talking to make his point and think about exactly what you’re saying and then come up with a response.
Translate
That’s not entirely the case: Google also introduced something like this in Gemini some time ago. ChatGPT can now answer before you finish talking. Hopefully he won’t talk over you all the time, but it’s not really meant for a normal conversation either. The idea is that you can use this to speak in English, for example, and ChatGPT will then translate what you say. This can be done in Spanish or Hindi, for example, while he continues to listen to what you say in order to translate it. As humans, we also think in the background and that is when GPT-Live briefly sends the question to GPT-5.5. In the meantime, he just keeps talking to you, while 5.5 actually does the heavy lifting. You can now also notice it yourself, because it is now rolling out to all users, including free users, although you may also have it from tomorrow. The rollout would not take much longer. GPT 5.6 is also expected today, but it is still unclear what it entails.