OpenAI adds voice control to desktop ChatGPT for complex tasks

OpenAI recently updated its ChatGPT desktop app with ChatGPT Voice, allowing users to talk to the AI and control it to do tasks. This feature uses the new ChatGPT-Live voice model and works with ChatGPT Work and Codex. It can also operate the computer, like opening websites and apps. On macOS, with Appshots, ChatGPT can read what's on the screen, including alt-text. Unlike the mobile version which focused on smoother conversations, the desktop version is more powerful. Users can give multi-step commands, and ChatGPT will ask for info or confirmation when needed. In a demo, a developer used one voice command to make ChatGPT create a code thread, submit a pull request, and find a bug. Also, remote access from iOS lets you use voice in Codex. Meanwhile, Anthropic updated Claude's voice mode too, using Opus, Sonnet, and Haiku models to handle tasks in apps like Gmail and Calendar.