OpenAI brings ChatGPT Voice to desktop with GPT-Live for multi-step commands
Now you can control Codex and agents on your desktop with just your voice.
OpenAI has updated its ChatGPT desktop app with ChatGPT Voice, leveraging the new GPT-Live voice models. This feature lets users speak complex, multi-step commands to control AI agents within ChatGPT Work and Codex. For example, a developer can ask ChatGPT to create a new thread, make a pull request, and find the root cause of a bug in one utterance. On macOS, Appshots gives the app permission to access screen content including alt-text, enabling it to assist with visual interfaces. The voice mode handles interruptions smoothly and can speak, listen, and coordinate work simultaneously, making it far more capable than the earlier smartphone version.
Anthropic has also updated Claude’s voice mode, which now taps Opus, Sonnet, and Haiku models to complete tasks across apps like Gmail, Calendar, Slack, Notion, and Canva. Both updates signal a shift toward hands-free, conversational AI control of professional workflows. For users, this means faster execution of routine and complex tasks without switching between tools or typing. The desktop voice agents can look up websites, interact with software, and chain multiple actions, effectively turning a voice command into a full workflow automation.
- Voice commands on desktop can orchestrate multiple AI agents and complex workflows like making pull requests or debugging.
- macOS Appshots integration allows ChatGPT to see and interact with on-screen content (including alt-text).
- Anthropic simultaneously upgraded Claude's voice to work across Gmail, Calendar, Slack, Notion, and Canva.
Why It Matters
Voice-controlled AI agents on desktop streamline complex workflows, making hands-free multitasking a reality for professionals.