OpenAI’s new voice AI can listen, think, and talk back in 70+ languages

OpenAI launched three new audio models in its Realtime API for voice apps.

GPT-Realtime-2 handles live reasoning and tool use; GPT-Realtime-Translate supports speech translation in 70+ languages.

GPT-Realtime-Whisper enables streaming transcription for captions, meeting notes, and other speech-to-text uses.

Early developer use cases include voice assistants, travel and home-search workflows, and multilingual product experiences.