To use ChatGPT Voice, open the app on mobile or go to chatgpt.com, tap the waveform icon in the message box, and pick a voice the first time you use it — after that it's one tap and you're talking. What changed recently isn't the button, it's what's on the other end of it: since July 8, 2026, that conversation runs on GPT-Live, a full-duplex model that can listen and talk at the same time instead of waiting its turn.
Short answer: Open ChatGPT on mobile or at chatgpt.com, tap the waveform icon in the message box, and pick a voice the first time you use it. As of July 2026, voice conversations run on GPT-Live, which listens and speaks simultaneously instead of taking turns. It's free for every logged-in user; paid plans get the fuller version of the model and, on desktop, hands-free control of ChatGPT itself.
I've been talking to ChatGPT Voice most days for the past few weeks — on my phone during my commute, and in the ChatGPT desktop app since it picked up voice controls on July 23, 2026 — specifically to see what changed since GPT-Live replaced the older Advanced Voice Mode. Here's what actually happens at each step, what's free versus gated behind a subscription, and the mistakes that waste a session.
What you'll need
Related: how to use ChatGPT for health questions.
Any ChatGPT account works, including the free tier — no credit card required to start a voice conversation. On mobile, you need the ChatGPT app itself (iOS or Android), signed in, with microphone permission granted; on a computer, chatgpt.com in a browser works the same way, and Plus, Pro, Business, Edu, and Enterprise users can also use voice inside the ChatGPT desktop app for macOS and Windows. Have a quiet-ish room and a real question ready — Voice works best when you're actually talking through something, not reading a script at it. If you're not sure whether you even want a paid plan yet, my guide to using ChatGPT for free covers exactly what the no-cost tier includes before you spend anything.
Step-by-step: how to use ChatGPT Voice
1. Open ChatGPT and tap the voice icon
On the mobile app or at chatgpt.com, look for the waveform icon inside the message box, usually to the right of the text field. Tap it to open a voice conversation.
2. Pick a voice the first time
The first time you start Voice, ChatGPT asks you to choose from nine named voices — Arbor, Breeze, Cove, Ember, Juniper, Maple, Sol, Spruce, and Vale. You can preview each one before deciding, and change it later in Settings if it doesn't fit.
3. Talk naturally — and interrupt whenever you want
Just speak like you would to a person. Since GPT-Live took over on July 8, 2026, the model can listen while it's still talking, so in my testing it handled interruptions and quick "wait, actually—" corrections far better than the old turn-based Advanced Voice Mode, which had to fully stop before it could hear you again.
4. Choose inline or Separate Mode
Voice can run right inside your chat window, or in Separate Mode, the full-screen animated orb view. Toggle between them in Settings → Voice → Separate Mode, depending on whether you want to see the transcript scroll by or just focus on the conversation.
5. Show it a photo or share your screen (subscribers)
On the mobile app, subscribers can tap the camera icon to share live video, or use the three-dot menu to share their screen or upload a photo mid-conversation. This part is gated: it's available to paid subscribers on iOS and Android, with only a limited preview on the free plan.
6. On desktop, let it drive while you talk
If you're on Plus, Pro, Business, Edu, or Enterprise, the ChatGPT desktop app for macOS and Windows now lets GPT-Live control your computer and coordinate agents running in ChatGPT Work or Codex, using only your voice — start a task, check on another thread, and steer both without touching the keyboard. OpenAI rolled this out globally on July 23, 2026.
7. Mute, end, and review the transcript
Tap the microphone icon to mute yourself without ending the session, and the exit icon to close it out. Every voice conversation leaves a written transcript in your chat history afterward, which is worth a quick skim if you talked through anything you need to act on.
Example prompts you can copy
These lean on what talking out loud does better than typing — pacing, interruption, and thinking through something messy:
- "I'm going to talk through my pitch out loud. Stop me the moment a sentence doesn't make sense and tell me why."
- "Walk me through changing a bike tire step by step, and don't move to the next step until I say I'm ready."
- "Quiz me on Spanish verb conjugations out loud, one at a time, and correct me right away if I get one wrong."
- "Look at what I'm holding through the camera and tell me if this plant needs water or more light." (subscriber camera feature)
- "Let's brainstorm names for [project] out loud — give me one at a time, and I'll say yes, no, or maybe."
Each one works because it needs the back-and-forth pacing or live camera that a text box can't really replicate.
Common mistakes to avoid
The one I made most often early on: expecting camera and screen sharing to work on the free plan — they're subscriber features on iOS and Android, and the free tier only gets a limited preview. Second is talking over ChatGPT out of habit from the old turn-based voice mode; GPT-Live is built to be interrupted, so there's no need to wait for a pause anymore. Third is forgetting to grant microphone permission the first time, which silently fails the session instead of prompting clearly. Fourth is treating a spoken answer as more trustworthy than a typed one — it isn't; double-check any number, date, or quote before you act on it, the same as you would with text chat. Fifth, and specific to the new desktop feature: trying to use voice computer-control on the free or Go plan and assuming it's broken — as of the July 23, 2026 rollout, that feature is limited to Plus, Pro, Business, Edu, and Enterprise.
ChatGPT Voice by plan
| Plan | Price | Voice model | Camera / screen share | Desktop computer control |
|---|---|---|---|---|
| Free | $0 | GPT-Live-1 mini, capped daily use | Limited preview only | No |
| Go | $8/month | GPT-Live-1 mini, capped daily use | Limited preview only | No |
| Plus | $20/month | Full GPT-Live-1 | Yes, iOS/Android | Yes |
| Pro | $200/month | Full GPT-Live-1, highest limits | Yes, iOS/Android | Yes |
| Business / Enterprise | $25/user/month (Business) or custom (Enterprise) | Full GPT-Live-1 | Yes, iOS/Android | Yes |
Free and Go both run on the smaller GPT-Live-1 mini model with a capped daily allowance before the conversation drops back to a lighter voice experience; every paid plan runs the full GPT-Live-1 model, according to OpenAI’s GPT-Live announcement.
Tools that make this easier
ChatGPT Voice is one feature inside a much bigger assistant — my guide to ChatGPT for work covers the plan and privacy setup worth knowing if you're using any of this for a job, and how to use ChatGPT agent mode covers the multi-step research side that voice now hands off to on desktop. If you're mostly on your phone, how to use ChatGPT on iPhone covers the mobile app in more depth, and how to use ChatGPT on a MacBook is worth a look before you try the new desktop voice-control feature. Comparing voice assistants across providers is also worth doing before you settle on one: how to use Gemini Live covers Google's equivalent, and how to use Claude AI covers Anthropic's setup, though Claude doesn't currently offer a comparable spoken-conversation mode. If your actual work is long-form writing rather than talking things through, a dedicated tool wins there — my ranked best AI writing tools guide and hands-on Jasper review cover the ones worth paying for.
My take
GPT-Live is a real upgrade, not a rebrand — being able to interrupt it mid-sentence and have it actually respond to the interruption, instead of finishing its thought first, is the difference that shows up in every session. It's free to try for anyone logged in, so there's no reason not to tap the icon once. The gap between plans is real, though: free and Go get the smaller model with a capped daily allowance and no camera or screen sharing, while Plus and up get the full model, live video, and now hands-free control of the desktop app itself. If you're only ever doing quick spoken questions, free is fine. If you want it running alongside real work — reading your screen, watching through the camera, or driving Codex tasks by voice — that's where the subscription earns its keep.
Frequently Asked Questions
Is ChatGPT Voice free?
Yes. Voice conversations are available to every logged-in user at no cost, running on the smaller GPT-Live-1 mini model with a capped daily allowance. Paid plans — Go through Enterprise — unlock the full GPT-Live-1 model, plus camera, screen sharing, and desktop computer control on Plus and above.
How long does it take to learn how to use ChatGPT Voice?
Starting a conversation takes seconds — tap the waveform icon and pick a voice. Getting comfortable with muting, switching to Separate Mode, and (for subscribers) sharing your camera or screen usually takes one or two real sessions.
What is the easiest way to use ChatGPT Voice?
Open the app, tap the voice icon, and just talk like you would to a person — interrupt it whenever you want instead of waiting for a pause, since GPT-Live is built to handle that now.
Can I use ChatGPT Voice on my computer?
Yes. At chatgpt.com in any browser, voice works the same as on mobile. Plus, Pro, Business, Edu, and Enterprise users can also use voice inside the ChatGPT desktop app for macOS and Windows, where it can control your computer and coordinate ChatGPT Work or Codex agents by voice — a feature OpenAI rolled out globally on July 23, 2026.
Does ChatGPT Voice work with the camera or screen sharing?
Partially, and only for subscribers. Live video and screen sharing are available on the iOS and Android apps for paid plans; the free tier gets a limited preview rather than full access.