How to use ChatGPT Voice: Live, Advanced and Standard modes, limits and settings

ChatGPT Voice lets you talk to ChatGPT and hear it answer, in a free-flowing conversation rather than a typed exchange. It works on the web, iPhone and Android, and since OpenAI’s GPT-Live models arrived in July 2026 it can listen and speak at the same time, so you can interrupt it the way you would a person. This guide covers the three Voice options, the daily limits on each plan, how to start, and the settings and privacy controls that matter.

ChatGPT Voice at a glance

The three Voice options

Open Settings, then Voice, to see which of these your account offers. Your plan, workspace settings, region, app version and any parental controls decide what appears.

How to start a Voice conversation

On iPhone or Android

1. Open ChatGPT and select the Voice icon in the message bar.
2. Allow microphone access if prompted.
3. If it is your first time, choose a voice.
4. Start speaking. Use the microphone control to mute or unmute, and the exit control to finish.

On the web

1. Go to ChatGPT and select the Voice icon in the prompt window.
2. Allow your browser to use the microphone if asked.
3. Start speaking. You can mute, unmute and exit the same way.

Voice works inside a normal chat, so the conversation appears as text while you talk and you can scroll back through earlier messages. Only one Voice conversation can run at a time.

Daily limits by plan

Live usage is measured over a rolling 24-hour period, and ChatGPT tells you when you reach the limit. OpenAI simplified the limits on September 9, 2026.

Plus and Pro no longer switch to GPT-Live mini when they reach a Voice limit. The separate Instant, Medium and High Voice intelligence levels have been retired.

Use text, images and better models in Voice

• Text and images: while Live is active, use the add button in the message bar to attach an image, or type instead of speaking. ChatGPT replies in Voice without starting a new chat.
• Harder questions: since September 9, Voice can use GPT-5.6 or GPT-6 Astra when it needs to search or reason. Choose the model and reasoning effort with the same controls as text chat; available models depend on your plan. For the difference between them, see our guide to ChatGPT models.
• What Live can’t do yet: find or add files from your ChatGPT Library, share video or your screen, or use connected apps and plugins.

Voices, language and style

Under Settings, then Voice, you can change:

• Voice: Arbor (easygoing and versatile), Breeze (animated and earnest), Cove (composed and direct), Ember (confident and optimistic), Juniper (open and upbeat), Maple (cheerful and candid), Sol (savvy and relaxed), Spruce (calm and affirming) and Vale (bright and inquisitive). In Brazil, Viola and Rio are also offered.
• Language: choose the one you speak most often for better recognition, or ask ChatGPT during a conversation to speak another one. Changing the voice mid-call starts a new voice call in the same chat.
• Style: the preset ChatGPT personalities don’t apply to Live, but you can ask it to change its tone, pace or style during a conversation. You can ask it to speak faster or slower, though there are no precise speed controls yet.

Settings worth turning on

• Background conversations: keep talking while you use other apps or your phone is locked. It ends when you end it, force-close the app, reach a limit or hit the maximum session length.
• Start with Voice: on supported mobile versions, opening a new conversation starts Voice automatically.
• CarPlay: ChatGPT works in Apple CarPlay on supported iPhones, where you can start Voice, continue a recent chat or begin a conversation in a project. Only use your phone when the law and conditions allow, and set the app up before you drive.

Get better results

• Use headphones and a quiet spot. On iPhone, open Control Center during a call, choose Mic Mode and turn on Voice Isolation.
• Ask it to wait. At the start you can say, “Wait until I ask you to respond,” if you want to think out loud.
• Give exact details for date, time or location questions. Voice uses your device or browser time zone to understand words like “today”, so add the date, time zone or place if an answer seems off.
• Keep it one-to-one. Live is designed for one person, and it may reply when people are talking to each other.
• Check important answers. ChatGPT can make mistakes, and transcripts are not verbatim records.

Privacy and your recordings

• Retention: audio clips from Live and Advanced conversations, and video clips from Advanced, are stored with the transcript in your chat history and kept for 30 days.
• Deleting: when you delete a chat, its audio and video clips are deleted within 30 days unless OpenAI needs to keep them for security, safety or legal reasons. Archiving doesn’t delete anything.
• Standard mode: audio is deleted after transcription, even if transcription fails.
• Training: OpenAI doesn’t use audio or video clips for training unless you choose to share them. If Improve the model for everyone is on, it may use transcripts and other files, depending on your plan and settings. Free, Plus and Pro users in personal workspaces can opt in to sharing clips; Business, Enterprise and Edu workspaces can’t share them.

Voice for work

Business, Enterprise, Edu and Healthcare workspaces can enable Voice in Chat, powered by GPT-Live, on the web, iOS and Android. Voice in Desktop, in the ChatGPT desktop app on macOS and Windows, lets you start tasks, check progress and coordinate agents by voice; it uses your Codex usage budget at $0.05 a minute, or 1.25 credits a minute on credit-based plans.

ChatGPT Voice versus other voice assistants

Google’s answer is Gemini Live, now powered by Gemini 3.8 Live Extended Thinking — see our report on Gemini 3.8 Live. Apple’s rebuilt assistant is covered in Siri AI in iOS 27, and Amazon’s in our Siri vs Alexa+ comparison. For ChatGPT’s plans, models and limits, read ChatGPT Go vs Plus or the ChatGPT profile.