Transcribe an audio file with the OpenAI /v1/audio/transcriptions endpoint, including timestamps or diarization

Documented steps

1. Prepare audio in a supported format: flac, mp3, mp4, mpeg, mpga, m4a, wav or webm. The speech-to-text guide documents a 25 MB upload cap — chunk longer audio yourself.
2. Pick a model. The reference currently lists: whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe, gpt-4o-mini-transcribe-2025-12-15, gpt-transcribe and gpt-4o-transcribe-diarize.
3. POST multipart/form-data to https://api.openai.com/v1/audio/transcriptions with header Authorization: Bearer $OPENAI_API_KEY.
4. Add form field file = the binary audio (an actual file part, not a path string) and model = the chosen model ID.
5. Choose response_format against the per-model support table: json works on all listed models; text is documented for whisper-1 and gpt-4o-transcribe-diarize; verbose_json, srt and vtt are whisper-1 only; diarized_json is gpt-4o-transcribe-diarize only.
6. For word or segment timestamps use whisper-1 with response_format=verbose_json plus timestamp_granularities[]=word and/or segment. The reference states this option is not available for gpt-4o-transcribe-diarize.
7. Optionally add language (ISO-639-1, e.g. en) to cut latency and improve accuracy, and prompt to steer vocabulary or style.
8. Optionally add temperature (0-1, default 0).
9. For incremental output set stream=true; the reference states streaming is not supported for whisper-1 and is ignored there. Streamed responses arrive as SSE transcript delta events.
10. For gpt-4o-transcribe-diarize with audio longer than 30 seconds, supply chunking_strategy (e.g. "auto"); this model also accepts known_speaker_names/known_speaker_references to label known speakers.
11. Parse the result: json/verbose_json expose text (verbose_json adds segments/words with timestamps), diarized_json exposes per-speaker segments, and text/srt/vtt return the raw string as the body.
12. For live microphone input use the separate Realtime transcription API instead of this endpoint.
13. Verify against the official reference: https://developers.openai.com/api/reference/resources/audio/subresources/transcriptions/methods/create and the guide https://developers.openai.com/api/docs/guides/speech-to-text