OpenAI just dropped three new voice models that could change how we interact with software. We’re talking GPT-Realtime-2, which has GPT-5-level reasoning and can handle complex requests while keeping conversations natural. Then there’s GPT-Realtime-Translate, a live translator that handles 70+ input languages and spits out 13 output languages in real-time. And finally, GPT-Realtime-Whisper, a streaming speech-to-text tool that transcribes as you speak. The big picture here is that voice is moving beyond simple call-and-response. Developers are building three types of experiences: voice-to-action (like telling an app to find you a house and schedule a tour), systems-to-voice (software proactively telling you about flight delays), and voice-to-voice (AI that goes back and forth with you naturally). Why it matters: This marks a shift from voice as a novelty to voice as a real productivity tool that can listen, reason, translate, and act simultaneously.