GPT - LIVE ADDS PROMPT - TO - VOICE AND REAL - TIME ACTION SELECTION : NO RECORDING REQUIRED

What happened: OpenAI expanded its GPT-Live developer API over two consecutive days, adding prompt-to-voice generation from a written description with no audio recording, plus a native connection between live voice sessions and the Decisions API. Key numbers: Speech output billed at $0.05 per minute. Why it matters: OpenAI states the changes lower the barrier to a branded voice persona and speed real-time action selection, two friction points it says have historically pushed teams toward third-party providers. OpenAI's Full Duplex Bench remains proprietary and not independently validated.

OpenAI expanded its GPT-Live developer API over two consecutive days, adding prompt-to-voice generation from a written description with no audio recording, plus a native connection between live voice sessions and the Decisions API.

OpenAI states the changes lower the barrier to a branded voice persona and speed real-time action selection, two friction points it says have historically pushed teams toward third-party providers. OpenAI's Full Duplex Bench remains proprietary and not independently validated.