OpenAI launches new voice intelligence features in its API

On Thursday, OpenAI introduced a suite of innovative voice intelligence capabilities to enhance its API, empowering developers to create applications that can engage in conversation, transcribe, and translate user interactions seamlessly. Among the new offerings is the GPT-Realtime-2 voice model, designed to deliver a more authentic vocal experience. This model, unlike its predecessor, integrates GPT-5-class reasoning, enabling it to handle complex user requests with greater sophistication. Another significant addition is GPT-Realtime-Translate, which provides real-time translation services that adapt fluidly to the flow of conversation. This feature supports over 70 input languages and 13 output languages, ensuring effective communication across diverse linguistic backgrounds. Additionally, the API now includes the GPT-Realtime-Whisper, a transcription tool that captures live speech and converts it to text as conversations unfold. OpenAI emphasizes that these models enhance the capabilities of voice interfaces, allowing them to listen, reason, translate, transcribe, and respond dynamically. These advancements are particularly beneficial for businesses looking to enhance customer service, but their applications extend to education, media, events, and creative platforms as well. However, OpenAI acknowledges the potential for misuse of these technologies and has implemented safeguards. These include built-in mechanisms to detect and halt conversations that violate their guidelines on harmful content. All these new voice models are available within OpenAI’s Realtime API, with translation and transcription billed by the minute, while the GPT-Realtime-2 model is charged based on token usage.