Gpt-4o-transcribe-diarize returns “chunking_strategy is required” — any working example or schema?

API
1

Hi everyone,

I’m testing the new gpt-4o-transcribe-diarize model for diarization (speaker separation) using the official Python SDK, but the API keeps returning this error:

Error code: 400 - {
“error”: {
“message”: “chunking_strategy is required for diarization models”,
“type”: “invalid_request_error”,
“param”: “chunking_strategy”,
“code”: “invalid_value”
}
}

Here’s the minimal Python example I’m using:

from openai import OpenAI
import os

client = OpenAI(api_key=os.getenv(“OPENAI_API_KEY”))

with open(“test.mp3”, “rb”) as audio_file:
transcript = client.audio.transcriptions.create(
model=“gpt-4o-transcribe-diarize”,
file=audio_file,
chunking_strategy={
“type”: “auto”
}
)

print(transcript)

The same code works perfectly with gpt-4o-transcribe and whisper-1,

so I assume the issue is specific to the diarization model.

Questions:

  1. Is there any official format or schema for chunking_strategy?

  2. Does the openai Python SDK (>=1.50) currently support diarization?

  3. Any working example or preview doc available for gpt-4o-transcribe-diarize?

I’m testing this on macOS with Python 3.11 and openai==1.51.0.

Goal: transcribe and separate speakers for customer service calls (agent / client).

Thanks in advance :folded_hands:

— Big Mike

2 Likes
2

There is an example here in the docs.

Most likely your python package is outdated.

Try running pip install --upgrade openai and remove the chunking_strategy parameter, as the sdk defaults to “auto”.

This one works fine to me:

audio_filename="yourfile.mp3"
audio_file = open(audio_filename, "rb")
transcript = client.audio.transcriptions.create(
    file=audio_file,
    model="gpt-4o-transcribe-diarize",
    response_format="diarized_json",
)
print(transcript.text, transcript.to_dict())
2 Likes
3

Thanks for the suggestion!

I actually tried exactly the example from the docs:

audio_file = open("yourfile.mp3", "rb")
transcript = client.audio.transcriptions.create(
    file=audio_file,
    model="gpt-4o-transcribe-diarize",
    response_format="diarized_json"
)
print(transcript.text)

But I still get this error:

Error code: 400 - {'error': {'message': 'chunking_strategy is required for diarization models', 'type': 'invalid_request_error', 'param': 'chunking_strategy', 'code': 'invalid_value'}}

I’m on Python SDK 2.6.1, so it seems that this version still requires chunking_strategy to be explicitly set — even though the docs say it should default to “auto”.

4

How unusual, it works flawless for me.

When you specify the chunking_strategy parameter, does it still gives an error?

audio_file = open("yourfile.mp3", "rb")
transcript = client.audio.transcriptions.create(
    file=audio_file,
    model="gpt-4o-transcribe-diarize",
    response_format="diarized_json",
    chunking_strategy="auto",
)
print(transcript.text,transcript.to_dict())
2 Likes
5

Thanks so Much
i dont know why but your example works great
thanks again :folded_hands:

2 Likes