OpenAI Adds New Voice Intelligence Models to Its API

OpenAI has added new voice intelligence capabilities to its API, announced in a press release. The update introduces three models designed for live conversation, translation, and transcription.

The new GPT-Realtime-2 model uses GPT-5-class reasoning to manage complex spoken interactions while maintaining conversational flow. It supports features such as parallel tool calls, adjustable reasoning levels, and longer context windows. The model is built to handle tasks like scheduling, answering questions, or executing commands through natural speech.

GPT-Realtime-Translate enables live translation across more than 70 input languages and 13 output languages, maintaining conversational pace and fluency. It is designed for use in customer support, education, media, and multilingual events.

GPT-Realtime-Whisper provides streaming speech-to-text transcription for live scenarios, allowing captions, meeting notes, and other text outputs to appear as people speak. The Realtime API includes safeguards that monitor sessions and can halt conversations detected as violating harmful content policies.

All three models are available through the Realtime API. GPT-Realtime-2 is billed by token usage, while GPT-Realtime-Translate and GPT-Realtime-Whisper are billed per minute.