OpenAI has added ChatGPT Voice to its desktop app, allowing users to speak with AI agents while they complete coding and multi-step computer tasks. The update connects voice conversations with ChatGPT Work and Codex, including tools that can browse websites, use applications and request approval when needed.
The feature is powered by GPT-Live, OpenAI’s new family of voice models designed for smoother conversations and more reliable interruption handling. Unlike the earlier mobile voice experience, the desktop version can use available agent tools to take action rather than only respond verbally.
Voice Can Direct Multi-Step Work
Users can give Work or Codex a spoken instruction, interrupt the agent naturally and answer follow-up questions without switching to text. OpenAI demonstrated a developer asking Codex to create a thread, investigate the cause of a software bug and prepare a pull request through one voice command.
According to OpenAI’s official Work and Codex guidance, Voice can coordinate tasks using the permissions and tools available within the selected experience. Work is intended for longer assignments and finished deliverables, while Codex focuses on software development.
The system can use computer-control functions to interact with websites and supported applications. It can also pause and ask the user for clarification or permission before continuing with sensitive actions.
Appshots Provide Screen Context on macOS
On macOS, Appshots let users attach an application window to a Codex conversation through a keyboard shortcut. The feature shares a screenshot and available on-screen text, including accessibility information, so the agent can understand what the user is viewing.
OpenAI said this reduces the need to describe an interface or copy information manually into a prompt. Users remain responsible for choosing which window and information are shared.
Codex Tasks Can Continue From iPhone
Users can also monitor and direct supported Codex work from the ChatGPT mobile app. OpenAI’s remote access announcement says users can review progress, respond to questions, approve commands and continue active desktop threads from their phones.
The company introduced GPT-Live earlier in July as the voice technology now supporting ChatGPT Voice. The models are designed to manage natural turn-taking, interruptions and lower-latency spoken responses.
Featured image credits: Wikimedia Commons
For more stories like it, click the +Follow button at the top of this page to follow us.