OpenAI launches two versions of its speech transcription AI, GPT Transcribe, with customized support for real-time and asynchronous use

OpenAI has added two transcription models to its API: GPT-Live-Transcribe for low-latency live speech and GPT-Transcribe for batch and completed audio files.

The models are built to better handle context, noise, and real-world speech, and OpenAI says they beat earlier Whisper-based and live speech models in benchmarks.

The launch expands OpenAI’s speech AI stack for developers, making it easier to build accurate captions, transcription, and voice-driven product features.