OpenAI has released a suite of new voice intelligence features designed to help developers build applications that can transcribe, translate, and converse with users in real time. The company announced the updates on May 7, 2026, introducing three major capabilities: GPT-Realtime-2 for advanced voice conversations, GPT-Realtime-Translate for real-time language translation, and GPT-Realtime-Whisper for live speech-to-text transcription.

What Makes OpenAI's New Whisper Feature Different?

The new GPT-Realtime-Whisper capability represents a significant upgrade to OpenAI's existing transcription technology. Unlike previous versions, this tool captures speech-to-text conversions as interactions occur, enabling developers to build applications that transcribe conversations in real time rather than processing audio after the fact. This live transcription capability opens possibilities for customer service platforms, educational tools, and media applications that need instant text conversion of spoken content.

The broader voice intelligence suite works together to create what OpenAI describes as a shift from simple call-and-response interactions to more sophisticated voice interfaces. OpenAI stated that "together, the models we are launching move real-time audio from simple call-and-response toward voice interfaces that can actually do work: listen, reason, translate, transcribe, and take action as a conversation unfolds".

How Are These Tools Priced and Billed?

OpenAI has structured pricing differently depending on which feature developers use. GPT-Realtime-Translate and GPT-Realtime-Whisper are billed by the minute of usage, while the more advanced GPT-Realtime-2 voice model is billed based on token consumption, a measure of how much text the model processes. This tiered approach allows developers to choose tools based on their specific needs and budget constraints.

What Industries Could Benefit From These Features?

OpenAI has identified several sectors where these voice intelligence tools could have immediate impact. The company notes that the new features will assist with a wide array of areas, including:

• Customer Service: Companies can expand their support capabilities by automating transcription and translation of customer conversations.
• Education: Schools and online learning platforms can use real-time transcription to make lectures and discussions more accessible.
• Media and Events: News organizations and event producers can transcribe and translate content on the fly for broader audiences.
• Creator Platforms: Content creators can leverage these tools to reach international audiences through real-time translation and transcription.

What Safeguards Has OpenAI Built In?

Recognizing the potential for misuse, OpenAI has incorporated safety measures into its new voice models. The company has built guardrails designed to prevent the features from being used to create spam, fraud, or other forms of online abuse. Specifically, OpenAI embedded triggers in the system so that "conversations can be halted if they are detected as violating our harmful content guidelines". These safeguards represent an attempt to balance the utility of the tools with responsible deployment practices.

The GPT-Realtime-2 model itself represents a significant technical advancement. Unlike its predecessor, GPT-Realtime-1.5, this new version is built with GPT-5-class reasoning capabilities, which OpenAI says enables it to handle more complicated requests from users. This means the voice model can engage in more nuanced conversations and handle complex instructions, not just simple voice commands.

How Can Developers Access These New Tools?

All three new voice models are available through OpenAI's Realtime API (Application Programming Interface), which is the technical interface that developers use to integrate these features into their applications. The Realtime API consolidates these voice intelligence capabilities in one place, making it easier for developers to build sophisticated voice-enabled applications without having to piece together multiple separate tools.

The translation feature is particularly expansive in its language support. GPT-Realtime-Translate can comprehend more than 70 input languages, meaning it can understand speech in dozens of languages, and can output translations in 13 different languages. This broad language coverage makes it viable for global applications that serve international audiences.

These updates signal OpenAI's continued investment in voice as a primary interface for artificial intelligence applications. As voice-based interactions become more central to how people interact with technology, tools that can transcribe, translate, and reason about spoken language in real time are becoming increasingly valuable for developers building the next generation of AI-powered applications.