Friday, Jul 10 | --:--
Back to home

OpenAI Ships GPT-Realtime-2, Live Translate, and Realtime Whisper in the API

On May 7, 2026, OpenAI released a new generation of Realtime API voice models: GPT-Realtime-2 with GPT-5-class reasoning for live voice agents, GPT-Realtime-Translate for speech translation across 70+ input languages into 13 outputs, and GPT-Realtime-Whisper for streaming speech-to-text—all aimed at production voice apps rather than ChatGPT Voice alone.

Tech Insights Reporter 4 min read San Francisco, CA

TLDR

OpenAI on May 7, 2026 announced three new Realtime API voice models for developers: GPT‑Realtime‑2 (most intelligent speech-to-speech path yet, with GPT‑5-class reasoning for harder multi-turn voice agents), GPT‑Realtime‑Translate (live speech translation: 70+ input languages → 13 output languages), and GPT‑Realtime‑Whisper (streaming speech-to-text while the user is speaking). All ship in the Realtime API (Playground available); pricing includes $32/$64 per million audio input/output tokens for Realtime‑2 (with $0.40 cached input) and per-minute rates for Translate ($0.034/min) and Whisper ($0.017/min).

What shipped

From OpenAI’s primary post “Advancing voice intelligence with new models in the API” and TechCrunch/community notes:

Model Role Notes
GPT‑Realtime‑2 Live conversational voice agents Successor-class to Realtime‑1.5; better context and complex requests
GPT‑Realtime‑Translate Live translation 70+ → 13 languages
GPT‑Realtime‑Whisper Streaming STT Transcribe as speech happens
  • Surface: Realtime API (WebRTC/WebSocket/SIP-style voice stacks)—API-first, not a ChatGPT consumer Voice rebrand.
  • Positioning: Voice as an agent interface—reason, translate, and transcribe in the loop—not only turn-based chat TTS.

Why this story matters

May already had GPT‑5.5 Instant as the ChatGPT default (May 5). Realtime‑2 is the parallel story for builders: speech-to-speech agents with flagship-class reasoning and first-class translate/transcribe companions. That is infrastructure for phone agents, multilanguage support, and low-latency apps—distinct from consumer ChatGPT Voice.

Sources

Featured Image Alt Text

OpenAI Realtime API badge with voice waveform and three model chips: Realtime-2, Translate, Whisper.

Tags

OpenAI, Realtime API, GPT-Realtime-2, Voice, Translation, Whisper, Developers

Scroll to continue reading