Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
OpenAI is updating its voice AI technology to match text-based AI. Two new models, GPT-Live-1 and a small version, are now being rolled out to everyone ChatGPT Users.
The new binary model is based on the latest OpenAI models. You can choose from three different levels of intelligence, depending on how in-depth you want the answers to be. But the biggest update has to be that the voice sounds more human and more conversational, Atty Eleti, ChatGPT’s audio product lead, told reporters.
One of the biggest changes under the hood is a new process called Continuous Reaction. This framework allows the AI to receive information and produce output simultaneously – so in the case of audio, it can “listen” and “speak” at the same time. This is a change from Previous sound modes Who cannot respond until she finishes speaking.
It’s useful if you want to do live, simultaneous translation: you can speak in English and have ChatGPT translate what you say into Spanish, Hindi, or another language with only a short delay.
Part of the new architecture can also delegate or offload tasks to OpenAI frontier models. For example, it can answer a question you asked while searching for another question.
“When GPT-Live has to think through a question, it can delegate its complex logic and task to GPT-5.5, which can do things in parallel, and GPT-Live can still stay in conversation with the user,” said Kundan Kumar, research lead on the GPT-Live model. When this is done, he “seamlessly weaves” the answer into his response.
“This is exactly how humans interact with each other,” Ileti added. “We keep the conversation going while we think about the background.”
It also uses OpenAI The latest artificial intelligence design technologies To create some visual answers, when appropriate, “because sometimes the best answer is shown, and it’s not said out loud,” Eletti said. Think weather reports, sports scores, and other types of graphics.
This is an example of the graph that ChatGPT will create if you ask it to give you a weather report while in voice mode.
These new voice models are now available to all ChatGPT users, including free users and paying subscribers, on the mobile app and website. Don’t worry if you don’t see it right away; OpenAI says it may take a day or two to ensure everyone has access. This is separate from what was expected on Thursday Low chain GPT-5.6.
You’ll notice that one of the ways the AI voice tries to mimic human speech is by using filler words — those “ums,” “ers,” and “likes” we use while thinking our thoughts out loud. Those are new for this update. You can also interrupt it with fewer delays. You can make him sit quietly and listen, and only respond when his name is called as a wake-up call. Unlike Siri and Alexa, ChatGPT will actively listen until manually disabled.
OpenAI automatically opts you out of AI training using voice mode. Audio clips are stored for 30 days – so the AI has the context of previous conversations – and can be deleted.
However, having a more human-looking AI comes with risks. We have seen many cases (and Lawsuits) of damages Embodiment of artificial intelligence This can happen, especially for people struggling with their mental health. OpenAI says the new models expanded safeguards and performed better than previous models in “key safety areas” including self-harm, psychosis, violence and sexual content.
