ChatGPT Voice Launches with Full-Duplex Audio and Deep Integration with Work and Codex

ChatGPT Voice has officially launched. Its voice mode employs full-duplex audio technology, supporting real-time two-way communication between users and the system. In actual interaction, users do not need to wait for turn-based responses — they can directly interrupt, correct instructions, or change topics, and the system accurately captures intent and follows up on the conversation. The current version still has room for improvement in long-pause recognition and background noise handling, but the naturalness of voice expression and cross-terminal text generation coherence have been enhanced.

The voice mode has been integrated with workflow and automation tools Work and Codex. Users can verbally describe their goals to have the system execute tasks such as finding local product options, automatically opening webpages for comparison, planning navigation routes, and following through to payment. The system also supports cross-device coordination — users can issue commands outdoors via their phones and remotely control webpage modifications, document generation, or code writing tasks on office computers.

For safety reasons, the system requires manual user confirmation for sensitive operations such as payments, permission changes, or sending messages. Browser control speed and real-time visual interaction features are still being optimized, and the current version already possesses the basic capability to execute complex desktop tasks.