OpenAI's GPT-Transcribe and Live-Transcribe bring real-time STT to API
Two new models promise 50% lower latency and 95% accuracy across 99 languages...
Deep Dive
Two new transcription models, GPT-Transcribe and GPT-Live-Transcribe, are now available in the API.
Key Points
- GPT-Transcribe handles offline batch transcription with 95% accuracy across 99 languages
- GPT-Live-Transcribe offers real-time streaming with sub-200ms latency for voice interfaces
- Priced at $0.006 per minute, undercutting Whisper API and competitors like Deepgram
Why It Matters
Developers gain a unified, cost-effective speech-to-text solution directly within OpenAI's ecosystem, simplifying voice AI pipelines.