Voice mode
In short
Voice mode is ChatGPT's spoken conversation feature. You talk, it talks back, and the newest setting can listen and speak at the same time instead of waiting for you to stop.
- Also called
- ChatGPT Voice, Advanced Voice Mode, Voice chat
- Topic
- Using AI tools
- Level
- Beginner
- First described
- OpenAI, 25 September 2023, in "ChatGPT can now see, hear, and speak". Voice on phones used Whisper to transcribe speech and a separate text-to-speech model, built with professional voice actors, to answer2.
OpenAI's help centre calls it ChatGPT Voice: you speak, and you hear a spoken reply inside the same chat, so you can also read the words. There are three settings. Standard is turn by turn. Your speech is transcribed, then the model answers. Advanced is the previous real-time mode, and on a phone it is the one that can use the camera or share your screen. Live, the newest, is built to listen and speak at once. Paid plans run Live on a model called GPT-Live-1. Free accounts use a smaller model, GPT-Live-1 mini. What you see depends on your plan, workspace settings, region and app version1.
The first version was announced on 25 September 2023 and rolled out to Plus and Enterprise users on iPhone and Android over the following two weeks. Whisper transcribed the speech, and a separate text-to-speech model, recorded with professional voice actors, spoke the reply2. The May 2024 GPT-4o announcement describes that earlier voice mode as three models in a row, with a delay of about 2.8 seconds on GPT-3.5 and 5.4 seconds on GPT-4. GPT-4o was one model trained across text, images and audio, able to begin an audio reply in as little as 232 milliseconds, and 320 milliseconds on average3.
On 8 July 2026 OpenAI launched GPT-Live, which now powers ChatGPT Voice. It is full duplex: it keeps processing what you say while it is speaking, and decides many times a second whether to talk, wait or hand a harder question to another model. It can say "mhmm" while you are still going, or stay quiet if you pause to think. At launch, that background model was GPT-5.54.
Live is not every voice feature. The help page says it does not, at first, support video, screen sharing, connected apps or plugins. Those stay on Advanced. Use is counted over a rolling 24 hours, and OpenAI says the limits may change. The $200-a-month Pro plan is listed as unlimited GPT-Live-1. Go and Plus are listed at up to an hour of GPT-Live-1 on Instant and another hour on higher reasoning, plus two hours of the mini model. The free plan gets limited use of the mini model only1.
Example
You are cooking and cannot touch the phone. You start a voice chat and ask for a substitute for buttermilk. With Live you can talk over the answer if it starts listing things you do not have14.
Why it matters
Speaking is how many people use a chatbot when they cannot type. The reply is still a chatbot reply, so the text on screen is there to check.
Questions people ask
What is the difference between Live, Advanced and Standard voice?
Standard transcribes you, then answers. Advanced is the older real-time mode and, on the mobile apps, the one with video and screen sharing. Live, powered by GPT-Live, can listen and speak at the same time. Live does not start out with video, screen sharing or connected apps1.
Is ChatGPT voice mode free?
Free accounts get limited access to GPT-Live-1 mini over a rolling 24 hours. Paid plans get the larger GPT-Live-1 model, and the $200-a-month Pro plan is listed as unlimited. OpenAI says the limits may change, and not every plan or country sees every option1.
How did ChatGPT voice mode work before GPT-4o?
OpenAI's GPT-4o announcement describes it as three models in a row: one transcribed your speech, a language model wrote text, and a third spoke it. The delay was about 2.8 seconds with GPT-3.5 and 5.4 seconds with GPT-4. GPT-4o put audio in and audio out in one model3.
Related terms
Tools that use it
- ChatGPTOpenAI's general chatbot, the most used AI app in the UK and worldwide.
- GPT-LiveOpenAI voice model that can listen and speak at once, behind ChatGPT Voice