GPT Live Transcribe
OpenAI · USA · 2026
The same job as GPT Transcribe, done as the words arrive — and priced at 3.8 times as much per minute for that impatience.
GPT Live Transcribe was released on 28 July 2026 as the streaming half of OpenAI's new transcription pair. Where GPT Transcribe waits for a finished recording or a completed conversational turn, this model reads a live microphone, phone call or broadcast feed and emits transcript fragments as the speech arrives. It runs only inside realtime transcription sessions; there is no file endpoint. It inherits the useful parts of its sibling: a free-form description of what is being said, keyword hints for names and jargon, and a list of expected input languages rather than one. Its own addition is tunable latency — the developer decides how long the model may wait before committing to a fragment, trading a little delay for a little accuracy. The interesting number is the price. Live transcription costs $0.017 per minute against $0.0045 for the same audio handled as a file — 3.8 times more for the identical task, with immediacy the only difference. It lands at exactly the rate of gpt-realtime-whisper, the older streaming option it supersedes, so OpenAI cut the price of unhurried transcription by a quarter while leaving the live rate untouched. For anything that does not have to be read while it is being said, the file model is the cheaper answer by a wide margin.
▸Videos
No videos yet.
▸Reviews
No reviews yet. Be the first!