OpenAI Adds New Voice Models to the API

2026-05-09 · openai

OpenAI shipped a new batch of audio models for the API, aimed at making voice apps feel more natural and more actionable. The update includes reasoning-capable realtime voice, live translation, and streaming speech-to-text. This is the kind of release that can change what developers build, not just how they build it.

Key Features or Updates

GPT-Realtime-2 adds stronger conversational reasoning for voice interactions. GPT-Realtime-Translate and GPT-Realtime-Whisper round out the stack with live translation and live transcription.

Impact on Developers

This lowers the friction for building assistants that listen, understand, and act in real time. It also makes voice interfaces more viable for support, productivity, and multilingual workflows.

How to use it

Start by choosing the model that matches the job: reasoning, translation, or transcription. Then wire it into a low-latency app loop so the model can keep pace with the user.

Read Original Post →