OpenAI API Adds GPT-Realtime-2, Live Translation and Transcription Models

OpenAI has expanded its API with three new voice intelligence models for real time voice applications. The release includes GPT-Realtime-2, a conversational model designed for realistic spoken interactions with GPT-5 class reasoning, allowing it to handle more complex live conversations than earlier versions.

The company also introduced GPT-Realtime-Translate for real time spoken translation, with support for more than 70 input languages and 13 output languages. Following these, GPT-Realtime-Whisper adds live speech to text transcription, letting applications capture spoken interactions as they happen.

GPT-Realtime-2 is billed by token usage, while translation and transcription are billed by the minute. All three models are available through OpenAI’s Realtime API and can support use cases such as customer service, education, media, events, and creator platforms. OpenAI also noted potential risks around spam, fraud, and online abuse, adding safety triggers that can halt conversations if harmful content is detected.