Talking to Chat GPT just got a lot less awkward. The assistant’s voice capabilities have evolved from a 2023 system that used three separate models—speech‑to‑text, language generation, and text‑to‑speech—to the newer GPT‑Live model, which OpenAI describes as a "full‑duplex" architecture that lets it listen and speak simultaneously. This eliminates the previous gripe of the bot cutting you off mid‑sentence or mistaking a pause for the end of your turn. GPT‑Live also adds natural acknowledgments like "mhmm" or "yeah" while you speak and can offload complex reasoning to other capable models such as GPT‑5.5 in the background.

The voice model is available on Android, iOS, and the web, with Chat GPT Pro, Plus, and Go users getting the full GPT‑Live model and free‑plan users receiving a mini version. You can tap the waveform icon to start, choose intelligence levels (Instant, Medium, High) on paid plans, and keep the voice active even when switching apps. While screen and video sharing aren’t yet supported, you can revert to the older voice model in settings, pick a different voice, or set Chat GPT Voice as the default.

The result is a smoother, more conversational experience that feels less robotic and more responsive.