OpenAI's ChatGPT Voice Brings GPT-Live to Desktop for AI Agent Workflows

OpenAI has introduced ChatGPT Voice for its desktop app, expanding voice interactions across Chat, Work, and Codex so users can speak naturally while directing tasks, reviewing progress, and coordinating longer-running work without returning to the keyboard.

The feature is powered by GPT-Live, OpenAI's voice model family, and is available in the ChatGPT desktop app for ChatGPT Plus, Pro, Business, Edu, and Enterprise subscribers. Enterprise and Edu customers receive the feature through a two-week early-access period before broader availability within their workspaces. Users can also access ChatGPT Voice through Remote on iOS after pairing a phone with a desktop host, although availability depends on rollout status and workspace settings.

Unlike standard voice dictation, ChatGPT Voice is designed as a live conversation. Users can interrupt responses, ask follow-up questions, or change direction while ChatGPT continues working. The system can also launch separate threads for longer tasks, report on progress, summarize blockers, and accept additional spoken instructions without ending the conversation.

OpenAI positions the feature as a way to coordinate work across Chat, Work, and Codex from a single voice session. Example use cases include reviewing documents that need approval, starting coding tasks, checking active work, and requesting status updates while those tasks continue in the background. Any work initiated through Voice follows the same permissions as the Chat, Work, or Codex tasks it controls.

To begin using the feature, users must start a new conversation in voice mode by selecting Start new voice chat before sending their first message. Conversations that begin in another mode support voice dictation only. Existing voice conversations can be resumed later by reopening the chat and selecting Start voice chat again. Users can also assign a custom keyboard shortcut through the desktop app's Voice settings.

On macOS, ChatGPT Voice can incorporate screen context through a feature called Appshots. After enabling Screen context in the Voice settings, users can ask ChatGPT to "Take a look at this," allowing the assistant to capture the frontmost application window and use it as context during the conversation. OpenAI notes that an Appshot may include both the window image and accessible text, including content outside the visible scroll area. macOS may request Screen & System Audio Recording and Accessibility permissions before the feature can be used.

The company advises users not to share windows containing sensitive information because Appshots can capture more than what is currently visible. Organizations also have the option to disable screen context capabilities.

OpenAI distinguishes ChatGPT Voice from voice dictation by separating conversational interaction from speech-to-text input. Voice dictation converts spoken words into a prompt before it is sent, while ChatGPT Voice maintains an ongoing dialogue that supports natural turn-taking and task coordination.

The desktop app allows only one active voice conversation at a time. Voice usage is governed by a separate plan-based allowance measured in rolling five-hour windows, while any tasks launched through Voice continue to count against a user's Codex usage budget. If users cannot start a voice session, OpenAI recommends confirming that the feature is available for their plan and workspace, checking microphone permissions, ensuring another voice chat is not already active, and verifying Screen context settings and required permissions when Appshots are unavailable.

See more here:

This analysis is based on reporting from OpenAI.

Images courtesy of OpenAI.

This article was generated with AI assistance and reviewed for accuracy and quality.