- ChatGPT’s new voice mode is rolling out today to everybody, even Free users
- It allows for much more natural conversations and won’t interrupt if you stop talking
- You’ll be able to do simultaneous translation for the first time ever in ChatGPT
OpenAI has upgraded ChatGPT’s voice mode for everybody with two new models that are rolling out globally, starting today.
I listened to the new GPT-Live-1 model in a demo run by OpenAI, and it does sound much more natural than ChatGPT’s previous voice model.
The new model aims to address two particular problems with the existing ChatGPT voice mode. Firstly, the previous version just wasn’t as smart as the text version of ChatGPT. Secondly, it tended to interrupt too much. You notice this especially if you go quiet while you’re thinking of a reply — ChatGPT will often fill the gap by talking.
Sounding more intelligent
To get around the intelligence problem, the new model actually delegates harder questions to ChatGPT-5.5, then comes back with an answer. It will say things like “let me just check that for you” to let you know it’s doing this, which keeps the flow of conversation feeling natural and doesn’t make it seem like you have to wait too long for an answer.
It does the same thing with any answer it needs to look up on the web. So, for example, if you asked it when your team’s next match was in the World Cup, it would say something like “OK, let me check that” while looking it up using GPT-5.5, then give you the answer.
"Hey Chat"
OpenAI also demonstrated how the new ChatGPT voice mode is quite happy to stop talking and listen if you tell it to, without interrupting. You can simply ask it not to reply until you speak to it directly again, and it will wait.
Of course, this requires you to call it a name, which it doesn’t officially have. In the demonstration I saw, the OpenAI employee called it “Chat”, so he said “Hey Chat”, just like you would say “Hey Siri”. In practice that seems to work quite well.
Simultaneous translation
The final new feature of note is simultaneous translation. If you watch world leaders being briefed at places like the United Nations, you’ll see that they have an earpiece through which they receive a simultaneous translation in their own language of whatever the speaker is saying.
Now you can do this with ChatGPT. Say “I’d like you to simultaneously translate whatever I’m saying into [language]”, then start talking, and ChatGPT will provide a live translation as you speak. Seeing this in action was actually quite impressive and I could imagine it being very handy in several real world situations.
All major languages appear to be supported as well.
The future for AI
The new GPT-Live-1 models — there are two, the normal one and a mini version — will start rolling out for all users immediately, but it could take a few days to reach everybody. The smaller GPT-Live-1 mini model will be the default for Free users, while paid users get the full GPT-Live-1 model.
So far, ChatGPT’s voice mode has been a handy tool for when you need to use your hands for something and can’t type, but it’s never been good enough to become the standard way you interact with ChatGPT. Now it looks like OpenAI is trying to unlock the ability to use voice as the primary interface to AI, and it’s quite possible that this is the future OpenAI is aiming for. Today I think we've all just taken a step closer to it.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.
