ChatGPT Voice in Work: What's New, How to Access It, and Alternatives (2026)
On September 23, 2026, OpenAI brought ChatGPT Voice into ChatGPT Work on web and mobile and let Voice use plugins. In Work, you can ask it out loud to create documents, presentations, and spreadsheets, use connected apps, or work in a browser. Voice in Work requires both Voice and Work access.
This guide covers what changed, which Voice option does what, and how to get access on each plan. It also covers usage limits, data handling, and the alternatives worth comparing. Every detail comes from vendor help pages read on September 24, 2026.
What changed on September 23
The ChatGPT release notes list the update under September 23, 2026, with the title "Use plugins in Voice and get work done by speaking." Two changes shipped together.
First, the Live voice option now supports plugins on web, iOS, and Android. During a call, you can use "the plugins and connected apps available to your account" and follow written responses in the chat.
Second, ChatGPT Voice is now available inside Work. The release notes put it simply: "Ask it to create documents, presentations, and spreadsheets, use connected apps, or work in a browser." If a task is still running when you hang up, it can continue in text.
The release notes also set the access rule. "Voice in Work requires both Voice and Work access." Free and Go users can use Voice in Chat with the plugins their plan supports.
That second change is the one that matters for finished work. Inside Work, ChatGPT Voice becomes a way to start and steer a task that ends in a file.
The three Voice options
OpenAI's help center lists three options under Settings, then Voice. They behave differently, so it helps to know which one you are using.
| Option | How it works | What it adds |
|---|---|---|
| Live | Listens and speaks at the same time, so you can interrupt | Web search, memory, widgets, text and images, plugins and connected apps |
| Advanced | The previous real-time Voice experience | Video and screen sharing on supported mobile apps |
| Standard | Turn by turn; transcribes your speech before responding | A simpler, transcript-first flow |
The help center says Live is powered by GPT-Live-1 or GPT-Live-1 mini, depending on your plan. It also notes one gap: "Live does not support video or screen sharing." If you need to show your screen, you switch to Advanced on mobile.
The options available to you can depend on your plan, workspace settings, region, app version, and parental controls. The help center lists all five.
What Voice can do inside Work
OpenAI describes Work as "an agent designed for longer, multi-step work and finished deliverables." Its jobs include creating "a document, spreadsheet, presentation, report, or Site." ChatGPT Voice gives that agent a spoken front end.
In practice, three kinds of requests fit well:
- Start a deliverable. Describe the document or deck you need, the audience, and the sources to use. Work builds it while you follow the progress in the chat.
- Pull from connected apps. Ask Voice to use an email, calendar, or Slack plugin, then fold the result into the task.
- Hand off and continue later. End the call when you arrive at your desk. The help center says an unfinished task "can continue in text."
One detail changes how you use it. "On web and mobile, if an action needs your approval, Voice will prompt you to review it on screen." Spoken approval is not supported. So a voice session that sends an email or edits a file still needs a glance at the screen.
The desktop app works slightly differently. There, ChatGPT Voice can start tasks, check progress, and coordinate multiple agents through one conversation. It is available on macOS and Windows.
How to access ChatGPT Voice in Work
Access depends on two things: your plan and your workspace settings.
On personal plans, Work is available on web and mobile for eligible paid plans. The release notes add that Free and Go users can use Voice in Chat with the plugins their plan supports.
On Business, Enterprise, Edu, and Healthcare workspaces, the help center says ChatGPT Voice is available "subject to workspace settings." In managed workspaces, Voice in Work requires both Voice and Work access. Admins control Work through Workspace settings, then Permissions and roles.
To start a session on the web, open ChatGPT and choose Work. Select the Voice control and allow microphone access if your browser asks. On iOS and Android, select the Voice icon in the message bar.
One notice on the help center is worth knowing before you upgrade. It says new sign-ups and upgrades to the ChatGPT Pro $200 plan are temporarily paused. Existing subscriptions continue to renew as usual.
Usage limits and cost
In Chat, Live usage is measured over a rolling 24-hour period. The help center lists these allowances.
| Plan | Live usage in Chat |
|---|---|
| Free | Limited access to GPT-Live-1 mini; limits may change |
| Go | 3 hours with GPT-Live-1 mini |
| Plus | 3 hours with GPT-Live-1 |
| Pro $100 | 15 hours with GPT-Live-1 |
| Pro $200 | Unlimited GPT-Live-1 |
| Business Standard | 3 hours with GPT-Live-1, then 1.25 credits per minute |
| Business Premium | 15 hours with GPT-Live-1, then 1.25 credits per minute |
| Enterprise on usage-based USD pricing | $0.05 per minute |
ChatGPT Voice in Work follows your plan's Voice limits and billing. The help center adds that "Work tasks also count toward your normal Work usage, just as they do when you type." So a long document task costs the same whether you speak or type the request.
Voice in the desktop app has separate pricing and limits. Teams that use both surfaces should check the desktop pricing page before rolling it out widely.
Privacy and data controls
Every ChatGPT Voice session creates audio, so the retention rules matter.
The help center says audio clips from Live and Advanced conversations are stored with the chat transcript. "Clips are retained for 30 days." Deleting a chat also deletes its clips within 30 days, with exceptions for security, safety, or legal reasons.
With Standard, audio is transcribed first. OpenAI says it deletes the audio after transcription unless you chose to share audio for training.
Training on audio is opt-in. Free, Plus, and Pro users can share clips through Data Controls. Users cannot share audio or video clips from Business, Enterprise, or Edu workspaces.
Limits worth knowing
The help center is candid about where ChatGPT Voice falls short today. Five limits stand out for work use:
- Live cannot currently find or add files from your ChatGPT Library. You may still be able to attach a file manually.
- Live does not support video or screen sharing.
- It is "not yet optimized for conversations with multiple speakers," so the help center does not describe it as a way to capture meetings.
- "Voice transcripts are not verbatim records and may not exactly match what was said."
- You can have one Voice conversation at a time.
None of these block the main use case. They do mean that voice suits starting and steering a task, while reviewing the output still happens on a screen.
Use cases that make sense now
A few patterns fit the current feature set well:
- Brief a deck on the move. Describe the audience, the three points, and the source files while commuting. Review the draft when you sit down.
- Turn a calendar into a prep document. Ask Voice to check tomorrow's meetings and draft a one-page brief for each.
- Catch up, then act. Have it summarize email threads through a plugin, then start a reply draft or a status update in Work.
- Rehearse a talk. Use Live to practice an explanation out loud and ask for questions a listener might raise.
Each of these starts with ChatGPT Voice and ends with a file you still read before sending.
Alternatives worth knowing
Two other assistants offer voice conversations that can use connected apps. Their help pages describe different scopes.
| Assistant | Where voice works | Connected tools in voice | Screen or camera sharing |
|---|---|---|---|
| ChatGPT Voice (Live) | Web, iOS, Android, desktop | Plugins and connected apps on your account | Not in Live; Advanced on mobile |
| Gemini Live | Gemini mobile app | Some Connected Apps, such as Google Calendar, Keep, and Tasks | Camera and screen sharing |
| Claude voice mode | Mobile, desktop, and web | Gmail, Google Calendar, Google Docs, Slack, and more | Not described on the help page |
Gemini Live. Google's help page describes "a natural, free-flowing conversation with Gemini" that you can interrupt. It supports camera and screen sharing on mobile. It "can use some Connected Apps and other services," such as adding an event to your calendar, with support released gradually. The page says that for now, Gemini Live is not available in the Gemini web app.
Claude voice mode. Anthropic's help center calls it "a beta feature available to all plans" on mobile, desktop, and the web. In voice mode, Claude can use connected tools like Gmail, Google Calendar, Google Docs, and Slack. The page notes that Claude Fable is not currently available in voice mode.
If your goal is a finished document built by an agent, ChatGPT Voice in Work is the option whose help pages describe that job directly. If you mainly want to talk through a problem, all three are worth a try.
Where voice stops helping
Speaking a request is faster than typing it. Checking a chart, a table, or a number is not something you can do by ear. The help center makes the same point in its own way: transcripts are not verbatim, and approvals happen on screen.
That review step is where the source matters. When a deck includes figures from a spreadsheet, someone has to confirm that each number matches the file. Powerdrill Bloom builds charts, documents, and slides from the files you upload. Its homepage says: "Every number comes back with the page, the row and the figure behind it."
For a deck that starts from a topic rather than a file, our topic to PPT page covers that route. For other ChatGPT surfaces, see our explainer on what ChatGPT Work is and our overview of ChatGPT for Word.
How to get started
Start with one task you already do every week. A status update or a short brief is a good first test, because you know what the finished version should look like.
Open Work, start a Live session, and describe the deliverable in one go: the audience, the sources, and the format. Let the task run, then read the output on screen and correct it by voice or text. Keep a note of what you had to fix, so the next request can include it up front.
When the deliverable depends on your own data, you can try Powerdrill Bloom to turn a spreadsheet into a sourced chart or deck.
Frequently asked questions
What is ChatGPT Voice?
ChatGPT Voice lets you speak with ChatGPT and hear a spoken reply, while following the response as text in the chat. Its Live option can listen and speak at the same time. Since September 23, 2026, Live can also use plugins and work inside ChatGPT Work.
How do you use ChatGPT Voice in Work?
Open ChatGPT, choose Work, and select the Voice control to start a Live session. Then ask it to create a document, presentation, or spreadsheet, or to use a connected app. Voice in Work requires both Voice and Work access on your plan.
Is ChatGPT Voice free?
Free users get limited access to GPT-Live-1 mini in Chat, and OpenAI says those limits may change. Voice in Work needs Work access, which comes with eligible paid plans. Go and Plus include three hours of Live use in Chat per rolling 24 hours.
Can ChatGPT Voice use plugins?
Yes. Since September 23, 2026, the Live option can use the plugins and connected apps available to your account. If an action needs approval, you confirm it on screen, because spoken approval is not supported.
How long does OpenAI keep ChatGPT Voice audio?
OpenAI's help center says audio clips from Live and Advanced conversations are retained for 30 days with the chat transcript. Deleting the chat deletes its clips within 30 days, with some legal and safety exceptions. With Standard, audio is deleted after transcription.
Sources: ChatGPT release notes · ChatGPT Voice help article · ChatGPT Work and Codex help article · Talk naturally with Gemini Live · Claude Help Center, Use voice mode. Plans, limits, and availability read from vendor help pages on September 24, 2026.