OpenAI Unveils Advanced Voice Intelligence Features for Developers
OpenAI Unveils Advanced Voice Intelligence Features for Developers
OpenAI has introduced a new set of voice intelligence features to its API, expanding its real-time conversational AI capabilities for developers and businesses.
Announced on May 7, the update includes three new tools designed to improve voice interactions: GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper. These features aim to help applications communicate, translate, and transcribe conversations more naturally and efficiently.
The newly launched GPT‑Realtime‑2 is a next-generation voice model built with GPT-5-class reasoning. According to OpenAI, the model is designed to handle more complex user requests while delivering more realistic and human-like conversations compared to its predecessor, GPT-Realtime-1.5.
Meanwhile, GPT‑Realtime‑Translate offers live translation capabilities that can keep pace with natural conversations. The feature currently supports over 70 input languages and 13 output languages, allowing users to communicate across different languages in real time.
OpenAI also introduced GPT‑Realtime‑Whisper, a speech-to-text tool capable of transcribing conversations instantly as they happen.
In a statement, OpenAI said the new models are intended to move voice AI beyond basic call-and-response interactions by enabling systems to “listen, reason, translate, transcribe, and take action” during ongoing conversations.
The company believes the technology could benefit industries such as customer service, education, media, live events, and creator platforms. However, OpenAI acknowledged the possibility of misuse and said safeguards were added to prevent spam, fraud, and harmful content. The system can reportedly halt conversations that violate the company’s safety guidelines.
All three voice tools are now available through OpenAI’s Realtime API. GPT-Realtime-Translate and GPT-Realtime-Whisper are billed per minute, while GPT-Realtime-2 uses token-based pricing.
