Source: OpenAI Speech-to-Text Guide

OpenAI API guide to transcription: current and deprecated STT models, 25 MB file limit, prompting with keywords, diarization, streaming and realtime transcription.

Contents
  • What: official API documentation (Speech to text guide)
  • Author/publisher: OpenAI
  • Date: living document; read 2026-11-14
  • Original: Speech to text

Summary

Covers file transcription via /v1/audio/transcriptions, live transcription through Realtime sessions, and the migration away from Whisper-era models.

Key claims

  • Models (as of 2026-11): gpt-transcribe for recorded audio, gpt-live-transcribe for live audio via Realtime transcription, gpt-realtime-translate for speech translation.
  • whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe and gpt-4o-transcribe-diarize are deprecated with shutdown on 2027-02-26.
  • Files up to 25 MB; formats flac, mp3, mp4, mpeg, mpga, m4a, ogg, wav, webm. Split longer audio without cutting mid-sentence.
  • Context: prompt (free text), keywords (literal-term hints, which can occasionally cause unspoken terms to appear) and languages. whisper-1 prompts are limited to 224 tokens; a text model can post-process transcripts for terminology.
  • Word/segment timestamps (timestamp_granularities[]) only on the deprecated whisper-1.
  • Diarization: diarized_json with speaker, start, end; up to 4 known-speaker reference clips of 2-10 s; not available in Realtime sessions.
  • Streaming file transcription emits transcript.text.delta / .done events.

Feeds

Voice agents · Multimodal models

Suggest changes →