How to enable gpt-4o-transcribe-diarize for realtime transcription?

API
1

Realtime **transcription doc mentions: “**The model to use for transcription. Current options are whisper-1, gpt-4o-mini-transcribe, gpt-4o-mini-transcribe-2025-12-15, gpt-4o-transcribe, and gpt-4o-transcribe-diarize. Use gpt-4o-transcribe-diarize when you need diarization with speaker labels.” But api(/v1/realtime/client_secrets) return: Your organization does not have access to this transcription model. How to enable gpt-4o-transcribe-diarize?

1 Like
2

Hi @ch_l

Do you see the model gpt-4o-transcribe-diarize when you make a list models API call?

3

I got the same in realtime, “Your organization does not have access.”

From selecting the -diarize model right in the audio “playground”, as expected from the API reference, although this is low value to transcribe one side of a “chat” without a list of speakers and their audio samples.

Yes, I do have access and have used “diarize” with reference speakers on the audio/transcriptions API.

4

yes，it’s

{
        "id": "gpt-4o-transcribe-diarize",

        "object": "model",

        "created": 1750798887,

        "owned_by": "system"

}

I’m trying to use realtime transcription with diarize model. Is this okay?

5

I don’t need audio samples to truly identify who a voice belongs to; I only need to distinguish that different people are speaking different parts. Therefore, I think using the diarize model in realtime transcription fits my use case very well.