What is ChatGPT voice mode, and what changed in 2026?
ChatGPT voice mode is a spoken conversation inside a normal chat, and in 2026 it became a different product. OpenAI's ChatGPT Voice help page, read on 15 November 2026, says Voice "lets you speak with ChatGPT and hear a spoken response" and "works within a chat, so you can listen while following the response in text." The model behind the default option is GPT-Live, which can listen and speak at the same time.

The dates come from OpenAI's release notes. GPT-Live-1 arrived on 8 July 2026 for Free, Go, Plus and Pro on web, iOS and Android. A month later voice learned to take file uploads and work inside Projects. On 23 September it gained plugins, and voice reached Work on web and mobile. GPT-Live-1.1 followed on 15 October with "more natural accents across languages" and new voices for French, German, Hindi, Japanese, Korean and Mexican Spanish.
The older options are on their way out. OpenAI told users on 10 November 2026 that Advanced and Standard voice go away on 11 December. Its stated reason: "voice usage has nearly doubled" since July, and "virtually all people using voice in ChatGPT now use GPT-Live." Conversations you had in either mode stay in your history and can continue in text or voice. Since 9 September, Live can also hand a harder question to GPT-5.6 or GPT-6 Astra, which you pick with the same controls as text chat, per the same release notes.
How do you start ChatGPT voice mode?
You start it by selecting the Voice icon in the message bar. The steps below are from OpenAI's help page as of November 2026.
- On iOS or Android, open the ChatGPT app and select the Voice icon in the message bar. On web, go to chatgpt.com and select the Voice icon in the prompt window.
- Allow microphone access if your phone or browser asks.
- On your first Voice conversation, choose a preferred voice.
- Start speaking once Voice opens. Select the microphone control to mute or unmute, and the exit control to end the call.
Four settings under Settings > Voice are worth a minute. Voice picks the sound: Arbor, Breeze, Cove, Ember, Juniper, Maple, Sol, Spruce or Vale, plus extra Live voices in other languages and Bubbie in English. Language fixes a wrong guess about what you speak, and you can also just ask ChatGPT to switch languages mid-call. Background conversations lets a call run while you use other apps or your phone is locked. Start with Voice, on supported mobile versions, opens Voice by itself whenever you open a new or empty chat.
Changing the voice during a call starts a new voice call in the same chat. You can have one Voice conversation at a time. In CarPlay, ChatGPT can start a call, continue a recent or pinned chat, or open a conversation in a project, but OpenAI's page says to set it up before you drive.
Live, Advanced or Standard: which ChatGPT voice mode are you using?
You are almost certainly on Live, because OpenAI says virtually everyone is. The three options show up under Settings > Voice, and the table is compiled from OpenAI's voice page and its retirement notice as of November 2026.
| Option | How it works | Video and screen sharing | Status |
|---|---|---|---|
| Live | GPT-Live-1.1 or GPT-Live-1.1 Quip, depending on plan. Full duplex: listens and speaks at the same time, can use web search and memory | No | The default |
| Advanced | The previous real-time voice model, which waits for you to stop speaking before it answers | Yes, on iOS and Android for eligible subscribers | Retires 11 December 2026, except that video and screen sharing keep running on it |
| Standard | Three models chained: speech to text, a language model, then text to speech. Slower, and information is lost between voice and text | No | Retires 11 December 2026 |
Which one is on your phone depends on your plan, workspace, region and app version, so the options you see may differ. Voice with custom GPTs is the other casualty: OpenAI says it "will continue using Advanced voice until their deprecation on December 11, 2026."
If you need the camera or a screen share today, switch to Advanced in Settings. In Advanced you select the camera button during a call to share live video, or the more options menu (three dots) and then Share Screen. After 11 December, OpenAI's own wording is that these two features "will continue to be powered by Advanced voice until GPT-Live has this capability," and it gives no date for that.
What can you do with ChatGPT voice mode?
You can do most of what you do in text, aloud, with the answer also written in the chat. Five uses that OpenAI's release notes and help page support, each with a prompt you can say as written.
- Look something up while your hands are busy. Live uses web search. Say: "What time does the pharmacy on Main Street close tonight, and is the drive-through open?" Voice reads your device time zone to understand "tonight," so a wrong answer usually means a wrong clock or no location.
- Talk through a document. Since 7 August 2026 you can upload a file during a voice conversation, or use voice inside a Project that has its own sources and instructions. Say: "I just attached the lease. Read me anything about early termination and tell me what it costs."
- Think out loud without being cut off. Tell Live at the start: "Wait until I ask you to respond." OpenAI warns that long pauses or background sounds can still trigger an answer.
- Practice a language. Pick a Live voice in French, German, Hindi, Japanese, Korean or Mexican Spanish, then say: "Order dinner with me in Spanish and correct my grammar after each turn."
- Get work done by speaking. Live can use plugins and connected apps on your account. If an action needs approval, ChatGPT asks you to review it on screen, because "spoken approval is not supported." In Work on web and mobile, a task still running when you hang up continues in text.
You can type or attach an image in the same chat while Voice is active, using the add button in the message bar. You can also interrupt ChatGPT mid-sentence. Live can listen and speak at the same time, so a half-sentence correction works.
Is ChatGPT voice mode free, and how long does it last?
ChatGPT voice mode is free in limited form, and paid plans buy hours. OpenAI's help page gives Free "Limited access to GPT-Live-1.1 Quip" with no number. Go gets 3 hours a day on Quip, Plus 3 hours on GPT-Live-1.1, the $100 Pro tier 15 hours and the $200 tier no cap, all over a rolling 24 hours, as of November 2026. When the allowance runs out the call ends and you can carry on in text.
The full plan table, the reset arithmetic and the history of those caps live on our page about the ChatGPT voice limit, so this page doesn't repeat them. One rule here: voice in the ChatGPT desktop app is billed separately from voice in Chat, so a desktop call does not draw down your Chat hours.
Why is ChatGPT voice mode not working, and what can't it do?
Most voice trouble is one of five causes, and the fix is on OpenAI's help page for each. The page also lists what Live can't do, so check the second list before assuming something is broken.
- It interrupts you or stops talking. Use headphones, move somewhere quieter or turn the volume up. On an iPhone, open Control Center during the call, tap Mic Mode and turn on Voice Isolation.
- It answers the wrong date or place. Voice uses your device or browser time zone. Check it, or say the exact date, time zone or location.
- The call ended on its own. OpenAI says a call may end at a usage limit, at the maximum session length, or when a long conversation reaches its context limit. Start Voice again, or continue in text.
- It mishears you. Set your language under Settings > Voice > Language.
- Voice will not start. Allow microphone access when asked. Only one Voice conversation can run at a time, so end any other call first.
What Live can't do, per the same page as of November 2026: share video or a screen, apply the preset ChatGPT personalities, give you a precise playback-speed control (you can only ask it to speak faster or slower), or find files in your ChatGPT Library. It also isn't built for several people talking at once, and it may answer when two people speak to each other. The transcript added after a call is not a verbatim record.
On privacy, audio clips from Live and Advanced are stored with the transcript and kept for 30 days. Deleting the chat deletes its clips within 30 days. OpenAI trains on clips only if you opt in under Settings > Data Controls. If you want a recording turned into editable text instead of a conversation, use Dictation, which OpenAI describes as a separate feature. Our guide on whether ChatGPT can transcribe audio covers file uploads, their length cap and Record mode.
How does ChatGPT voice mode compare with Claude and Gemini?
All three assistants talk back, but they differ on video, tools and where they run. The table is compiled from each vendor's own help page on 15 November 2026, and the Gemini row comes from Google's Android page, so check yours if you are on iPhone or desktop.
| Assistant | How you start it | Model behind it | Video or screen | Tools in voice |
|---|---|---|---|---|
| ChatGPT | Voice icon in the message bar, on web, iOS and Android | GPT-Live-1.1, or GPT-Live-1.1 Quip on Free and Go | Not in Live; Advanced only, until it retires on 11 December | Web search, memory, plugins and connected apps on your account |
| Claude | Voice mode on web, Claude Desktop, iOS and Android | The same Claude models as text chat; follows your plan and starts from the model you last used | Not stated on the page | Gmail, Google Calendar, Google Docs and Slack; one connected tool on Free |
| Gemini | Gemini Live in the Gemini mobile app on Android | Not named on the help page | Camera and screen sharing in Live | Calendar, Keep, Tasks, Flights, Hotels, YouTube Music and more; Gems can't be used with Live |
Claude offers two ways to talk. Hands-free mode listens continuously and answers at natural pauses, and push-to-talk has you hold a button, which Anthropic suggests for noisy places. Switching between text and voice keeps the context. Voice mode isn't available in Claude Cowork or Claude Code, where only dictation works. Our page on Claude voice mode goes through it. Google's Android page says Gemini Live is not in the Gemini web app "for now."
Whizi, which publishes this page, has its own Voice mode and it loses to ChatGPT on time and on video. A Whizi voice session is capped at 15 minutes, there is no camera or screen share, and the monthly allowance is 80 voice minutes on Pro, which is under half of one day of ChatGPT Plus voice. Whizi voice runs on the Gemini Live API with a single preset voice, and saved memory is loaded when a session starts. What the Pro plan adds is Claude, GPT and Gemini in one app, with a first week at $1.99. The free plan gives 10 text messages a day with no card, and no voice. How voice works in Whizi has the details.
- Start a call from the Voice icon in the message bar on web, iOS or Android, and allow the microphone when asked
- Check Settings, then Voice, and stay on Live unless you need video or a screen share
- Plan around 11 December 2026: Advanced and Standard are gone from the picker that day
- Say the date, time zone and place when an answer depends on them
- Use headphones, and Voice Isolation on iPhone, if ChatGPT keeps cutting in
- Type in the same chat once the call ends or hits its limit
Frequently asked questions
Is ChatGPT voice mode free?
Yes, in limited form. OpenAI's help page gives Free users "Limited access to GPT-Live-1.1 Quip" and adds "Limits may change," with no hour count. Go and Plus get 3 hours a day and Pro $100 gets 15, as of November 2026. ChatGPT voice limit covers the rolling reset.
Can you use ChatGPT voice mode on desktop?
Partly. OpenAI lists Voice in Chat on web, iOS and Android. The ChatGPT desktop app for macOS and Windows has a separate voice feature for starting and steering Work and Codex tasks, with its own pricing and limits. On a computer, chatgpt.com in a browser is the way to talk in a normal chat.
What is the difference between ChatGPT Live and Advanced voice?
Live listens and speaks at the same time, while Advanced waits for you to stop talking before it answers. Live uses web search, memory and plugins and cannot share video or a screen. Advanced can on iOS and Android. Advanced retires on 11 December 2026 except for those two features.
Does ChatGPT voice mode work with custom GPTs?
Yes, for now. OpenAI says voice with custom GPTs keeps using Advanced voice until their deprecation on 11 December 2026. Custom GPTs explains how those are built and what replaces them.