Build more natural voice experiences with GPT‑Live‑1 in the API

Details

OpenAI introduced GPT-Live-1, a new voice model available through its API designed for real-time, full-duplex conversations. Unlike earlier turn-based voice implementations, GPT-Live-1 supports simultaneous speech and listening, enabling more natural dialogue flows without perceptible turn-taking delays. Key capabilities include stronger instruction following — a persistent weakness in earlier voice models — and custom voice support, allowing enterprises to deploy branded or persona-specific voices. The addition of telephony support is notable as it lowers the barrier for connecting AI voice agents directly to phone systems without additional middleware, opening practical paths for call center automation and IVR replacement. The source content is brief and does not include specific latency numbers, pricing details, or model architecture information. Developers interested in production deployment would need to consult OpenAI's API documentation for specifics on costs, rate limits, and supported telephony protocols. This release is directly relevant to teams building voice assistants, phone-based AI agents, customer service automation, and interactive voice response systems.