Voice Conversations with ChatGPT

ChatGPT Voice turns written exchanges into spoken conversations: you speak, ChatGPT replies out loud, and the transcript appears in the same chat thread. This isn’t just cosmetic. Voice makes the tool usable when typing is impossible or awkward—while walking, cooking, or between meetings. Since July 2026, the experience has relied on GPT-Live, a voice model that listens and speaks simultaneously.

Three voice options

In Settings > Voice, the OpenAI Help Center describes three options. Availability depends on your subscription plan, workspace, region, and app version:

• Live runs on GPT-Live-1.1 or GPT-Live-1.1 Quip, depending on your plan. It listens and speaks at the same time, making interruptions and follow-ups feel natural. It can use web search, memory, visual widgets, plugins, and connected apps, and it accepts both text and images in the same conversation. It does not support video or screen sharing.
• Advanced is the legacy real-time voice experience. Choose this option to share live video or your screen in the iOS and Android apps.
• Standard works turn-by-turn: your speech is transcribed first, then the response is generated.

To convert a simple recording into editable text without starting a conversation, use dictation instead.

Start a conversation

On mobile or the web, select the Voice icon in the input bar and grant microphone access if prompted. The first time you use it, pick a voice, then start speaking. During the conversation, you can mute your microphone or exit the voice session. Live usage limits depend on your plan over a rolling 24-hour period: limited access on the free tier, three hours on Go and Plus, and more on Pro and Business Premium. ChatGPT notifies you when you approach your limit.

What voice enables

With Live, the conversation feels like a phone call: you can interrupt, rephrase on the fly, or pause to think. When a question requires deeper research or reasoning, ChatGPT can switch to a more powerful model and continue the conversation while it works. If a plugin action requires your approval, you must confirm it on screen; voice approval is not supported.

One caution noted in the Help Center: voice uses your device’s time zone to interpret terms like “today” or “tomorrow.” For date-, time-, or location-sensitive questions, specify these details explicitly and verify critical information.

When voice shines

The benefit is obvious when you’re on the move: ask for an explanation, draft a message, or brainstorm without pulling out a keyboard. But the richest use case lies elsewhere: speaking forces you to articulate ideas differently than writing, which often helps unblock thinking. Dictate an oral draft, then ask for it to be formatted: “I’m going to describe my project, then you’ll summarize it into a 30-second pitch.”

Structure matters as much orally as it does in writing. “I have three questions about our launch. First question…” yields better results than a continuous stream. Context markers help too: “Let’s return to the budget,” or “Change the subject.” In very noisy environments, switch back to text rather than repeating the same sentence three times.

Switch between voice and text

Voice and text coexist in the same conversation: you can type when you can’t speak, attach an image during a call, or review the transcript later. This is valuable for capturing a spoken to-do list in writing or sharing a link that’s hard to pronounce. In Business, Enterprise, Edu, and Healthcare workspaces, voice availability depends on workspace settings; in ChatGPT Work, it also lets you start a task that continues in text after the call ends.

Try it yourself

A three-part session. Open Voice with the Live option. Introduce your role in 30 seconds, then ask: “Based on what I just told you, what are the three tasks in my day where you could help me the most?” Interrupt the response midway to clarify a point, then finish by asking for a written summary, which you’ll find in the conversation history.

Official sources: ChatGPT Voice · ChatGPT release notes (OpenAI Help Center) · Introducing GPT-Live (OpenAI)

Key takeaways

• Three options in Settings > Voice: Live (GPT-Live-1.1), Advanced for video and screen sharing, and Standard for turn-by-turn interaction
• Live listens and speaks simultaneously, uses search, memory, and plugins, but doesn’t support video
• Live usage is limited by plan over a rolling 24-hour window
• Approvals happen on screen; specify date and location for sensitive questions
• Voice and text coexist in the same conversation, including the transcript