OpenAI Adds Plugins and Mobile Work to ChatGPT Voice

OpenAI’s September 23 update gives ChatGPT Voice a more practical job than answering questions aloud. According to its ChatGPT release notes, Live Voice can now use supported plugins and connected apps on the web, iOS and Android. Voice also works in ChatGPT’s Work area on web and mobile, where eligible users can ask it to start document, spreadsheet, presentation and browser tasks.

The capabilities serve different users. Free and Go users can speak to supported plugins in Chat; people with access to both Work and Voice can use speech to start Work tasks. Account permissions and usage limits still apply.

The update gives people a way to instruct task-running software when typing is inconvenient, particularly on a phone. Whether they can finish a task without touching the screen is a separate question.

Voice Can Now Reach Connected Apps in Chat

Until this update, the useful parts of a connected app could be harder to reach during a voice conversation than during a typed one. OpenAI says plugin and connected-app support is now available in Live Voice across ChatGPT on the web, iOS and Android. A spoken request can call on a supported connection during a voice conversation.

Supported is the key qualifier. OpenAI has not promised that every plugin works by voice or that every account has the same connections. Speaking also does not grant ChatGPT access to an app the user has not authorized. The release notes say existing permissions and limits continue to apply.

Someone who already relies on connected apps can now make a request while walking or away from a keyboard, without switching to a typed chat first. The result will still depend on the connection, its access and the task requested. OpenAI has announced a new input path, not guaranteed success for every connected-app workflow.

Voice in Work Brings Task Creation to Phones

Work is the more consequential part of the update for people who use ChatGPT to produce deliverables. OpenAI says users with access to both Work and Voice can speak requests to create documents, presentations and spreadsheets, use connected apps, or work in a browser. Voice in Work is available on web and mobile.

In Work, a request to make a presentation is intended to produce an artifact or an ongoing task the user can inspect and continue. TechCrunch’s account of the rollout also describes examples such as drafting an email and summarizing Slack messages, subject to the relevant access.

OpenAI says unfinished Work can continue in text after a voice session. A spoken instruction may be enough to start a spreadsheet, while checking figures, revising slides or approving an action may be easier on screen. TechCrunch reports that users can start a conversation on the go and resume on desktop. The release notes’ narrower, explicit promise is that unfinished voice-initiated work can continue in text.

Mobile support changes when someone can start the work, though they may still prefer a keyboard and larger display for detailed review.

Access Depends on Both the Plan and the Feature

“ChatGPT Voice gets plugins” and “ChatGPT Voice gets Work” describe different entitlements. OpenAI identifies Free and Go users as able to use Voice in Chat with supported plugins. Voice in Work requires access to both Work and Voice, along with the permissions needed for the particular task.

TechCrunch specifically identifies Plus and Pro subscribers as users who can access mobile Work. That reporting does not establish identical access for every other plan or managed workspace. Feature availability, rollout status, workspace settings and connected-app permissions may affect what appears for an individual account.

The split depends on where the request is made and what it starts:

• Voice in Chat: Speak to ChatGPT and use a supported plugin or connected app where your account has access. OpenAI includes Free and Go users in this capability.
• Voice in Work: Speak to start a document, presentation, spreadsheet, connected-app or browser task. Your account must have both Work and Voice access.

Plugin access during an ordinary voice chat does not imply access to the Work tab. Work access, in turn, does not remove the separate requirement for Voice. OpenAI has not described the update as a universal, plan-independent release of mobile agents.

A Spoken Instruction Is Not App Authorization

Voice makes an action easier to request without changing the permissions behind it. If a task depends on a connected service, ChatGPT needs the access permitted by that connection. Workspace controls and existing limits still apply, according to OpenAI.

Approval screens deserve attention in a hands-free workflow. If ChatGPT asks a user to connect an account, grant access or confirm an action, the user should review what is being authorized. The initial spoken request does not mean every subsequent step has been approved.

The announcement does not establish that all Work and plugin tasks can be completed without looking at the device, or give a single rule for which actions require on-screen confirmation. Starting a task by voice is different from completing every permission, review and approval step by voice.

For work involving external accounts or editable files, users still need to check the destination, content and requested access before accepting an approval prompt. Voice reduces typing; it does not remove the need for oversight.

Mobile Is the New Part of an Existing Work Voice Story

OpenAI had already brought voice-driven tasks to Work on desktop. As TechCrunch notes, that earlier desktop availability followed its July launch of GPT-Live. September’s update extends Work voice use to mobile and adds plugin and connected-app support to Live Voice across web and phones.

Voice-controlled task execution itself is not new here. The expansion changes where users can begin those tasks and what a regular voice conversation can reach. For a feature previously tied to a desk, that broader availability may be valuable.

The announcement leaves open which supported connections work best by voice, how often a mobile task needs an approval screen, and how reliably work moves from spoken instructions into text-based revision. OpenAI’s release notes establish availability and intended workflows, not measured success rates. “Hands-free” may not describe every step.

Final Thoughts

Voice can now start Work tasks from a phone, and users can continue unfinished work in text when they need to review or revise it. Supported plugins also bring connected-app actions into ordinary voice chats for users who do not have Work access.

The practical limits remain account access, approval requirements and the quality of the resulting artifact. Starting a task aloud does not remove the need to check what ChatGPT does next.