GPT-Transcribe and GPT-Live-Transcribe

Two speech-to-text models in the OpenAI API, one for finished audio files and one for transcribing while someone is still talking.

GPT-Transcribe handles completed audio files and final transcripts of live turns, priced at $0.0045 per minute of audio
GPT-Live-Transcribe streams text as the audio arrives, with tunable latency, priced at $0.017 per minute
Both accept free-form context, keyword hints, and a list of expected input languages, which is how you get names, numbers, and industry terms spelled right
OpenAI says both handle accents, short phrases, specialized terminology, and loud background noise more accurately than what came before
They run on the audio transcription and realtime transcription endpoints only, not chat, batch, or fine-tuning

Turning recorded calls, interviews, and webinars into text, or putting live captions on audio as it plays.

Reach for GPT-Transcribe on audio you already have. Reach for GPT-Live-Transcribe when the text needs to appear before the speaker finishes.

Developers wiring transcription into a product, and teams pushing large volumes of recorded audio through an API instead of a desktop tool.

Staying informed is great. Staying ahead is better.

Pick a topic and leave with something you created.

Get new module drops and weekly AI strategy.