Comparison · OpenAI transcription

Choosing an OpenAI transcription model

GPT Transcribe offers three OpenAI speech-to-text routes. They differ in purpose more than in setup, so the right choice depends on what you will do with the transcript.

Choose a model and transcribe

In short

gpt-transcribe — default, OpenAI's current recommended start for recorded speech.
GPT-4o Mini Transcribe — lightweight, token-priced route for drafts.
GPT-4o Transcribe Diarize — speaker labels and timestamps.
All accept the same file types up to 25 MB.

Pick by job

Podcast episode for show notes → gpt-transcribe.
Quick voice memo to skim → GPT-4o Mini Transcribe.
Interview, meeting or panel → GPT-4o Transcribe Diarize.

Improving accuracy on any route

Add context with names, acronyms and specialist terms. Clean audio — close microphones, little background noise — matters more than the choice between routes.

Frequently asked questions

Which is the default?

gpt-transcribe.

How are credits calculated?

From the selected model and recording length; the exact quote appears before you submit.

Updated 2026-10-04. A focused OpenAI transcription workspace for recordings, interviews, meetings, and podcasts.