A new voice for ChatGPT has arrived: Check if you have it and what you can try with it

Although free and paid users get slightly different models – GPT-Live-1 mini and GPT-Live-1 – the updated system should appear on all ChatGPT accounts, bringing a number of new and exciting features.

How do you know you’ve received an upgrade?

The biggest indicator that you have received the new version will be the “Live” label at the top of voice conversations on mobile devices and within the drop-down menu on the web. By pressing that icon, you can still revert to the older voice models, at least for now.

There is another way of checking. On a mobile device, tap the menu button in the upper left corner, then the settings gear icon (Android) or your profile avatar (iOS), then select Voice > Model. On the web, click your profile avatar in the lower left corner, then Settings > Voice > Model.

What does GPT-Live bring?

The key innovation lies in the so-called “full-duplex” architecture, which enables the model to listen and speak at the same time. This solves one of the biggest limitations of previous versions, where the artificial intelligence had to wait for the user to stop talking before responding, which often led to awkward pauses. With the new system, users can freely interrupt the AI, and it can respond with confirmations like “mhm” and “yes” to show that it is actively listening, making the conversation more dynamic.

According to detailed testing conducted by TechRadar, it’s this “duplex” functionality that’s the biggest upgrade you should try first. Try to keep talking even after ChatGPT has started responding to you and you will notice how the system follows you. Then try interrupting him in the middle of a sentence and you’ll see how he adjusts his answer. The experience is now closer to the interaction from Spike Jonze’s 2013 sci-fi film Her.

New possibilities that are worth exploring

Another big innovation is real-time translation. Simply ask ChatGPT in voice mode to translate something into a foreign language as you say it. You can say sentences in Croatian or English, and ChatGPT will translate them without hesitation. Although perhaps not ideal for language learning, this function impressively demonstrates the speed and capabilities of the new model.

The third thing you can try is changing your voice and level of intelligence. You can do this via the slide icon in the upper right corner of the voice conversation. The voice options are the same as before, but now you can choose between three levels of intelligence: Instant, Medium, and High. Use Instant for the fastest answers, High for the best quality, and Medium as a compromise between speed and accuracy.

Finally, try asking questions that require visual responses. OpenAI has added a series of visual cards, so now you can get graphic displays of sports results, weather forecasts or map locations on your screen. This is possible because the model can delegate more complex tasks that require web browsing or advanced locking to a more powerful model, such as GPT-5.5, while continuing to talk to you.

Initial problems and user reactions

Although the conversation experience with the new model has been greatly improved and feels more natural than ever, some users have reported initial problems. On Reddit, OpenAI representative Atty Eleti answers user questions and acknowledges the existence of several bugs. One of the main problems is related to ChatGPT’s memory, which in some cases becomes unavailable in voice mode. OpenAI said that it is “a problem that is being actively investigated”.

Problems have also been reported with the pronunciation of foreign languages, which sometimes sound like someone with a strong English accent. This problem is also recognized and its improvement is expected over time. Despiteč despite these initial difficulties, the introduction of the new feature goes relatively smoothly, and most users are impressed by the naturalness of the conversation. For those who frequently use voice mode, this is definitely a big step forward.