If you enjoy talking with ChatGPT, the experience has become even more natural. OpenAI’s latest voice model can carry on conversations much like a real person. It can continue a discussion after being interrupted, slow down its speech when asked and respond without awkward pauses.
One of the biggest improvements is that ChatGPT no longer waits for you to finish speaking before processing what you say. Instead, it can understand speech in real time, respond naturally during the conversation and even use brief acknowledgments such as «uh-huh» or «I see» to show that it is actively listening.
If the AI’s voice feels too fast, you can simply ask it to slow down. ChatGPT will adjust its speaking pace on the fly without interrupting the conversation.
The updated voice mode also integrates with other ChatGPT capabilities. During a conversation, it can search the web for information, display weather forecasts, sports scores and stock prices, work with images and files, and even provide real-time speech translation.
OpenAI has also strengthened the system’s safety features. If ChatGPT detects a potentially harmful conversation, it may adjust its response, display a warning, or, in some cases, end the conversation. For child and teen accounts, parents can disable voice mode and receive notifications if the AI detects potentially concerning topics during a conversation.
The new voice mode is being rolled out gradually on ChatGPT for the web, as well as on Android and iOS. Users on the Go, Plus and Pro plans will have access to the full GPT-Live-1 model, while Free users will receive the lighter GPT-Live-1 mini version.
