GPT-Realtime-Whisper brings streaming speech to text

This is the unglamorous part of voice AI that actually matters: if transcription lags, the whole experience feels broken. GPT-Realtime-Whisper makes the Realtime stack more useful for production workflows by shrinking the delay between speech and text.

• –Live STT is the substrate for captions, note-taking, support triage, and voice agents that need continuous understanding
• –Streaming transcripts unlock partial results earlier, which matters more than perfect end-state text in real-time products
• –OpenAI is pricing it at $0.017/min, which signals this is meant for high-volume operational use, not demos
• –The release reinforces the idea that voice stacks are becoming modular: reasoning, translation, and transcription are now separate building blocks
• –For developers, the main win is UX: less waiting, fewer “please hold” moments, and more natural conversation flow