Error in Transcribing Audio File with GPT-4o-Transcribe Model

Question

When testing the deployment of a chatbot to the public, an error occurs both in the playground and in the API:

Error: The server had an error processing your request. Sorry about that! You can retry your request, or contact us through an Azure support request at: https://go.microsoft.com/fwlink/?linkid=2213926 if you keep seeing this.

The GPT-4o-transcribe model was deployed for transcription, and an API request was made for the model.

What steps should be taken to resolve this issue?

Answer 1

Hi Amber Cubacub

Here are some steps:

1. Ensure that the audio file is in a supported format. Only WAV format is supported for transcription.
2. Verify that the request headers, especially Content-Type and Content-Length, are set correctly.
3. emporary network issues could also lead to failures. If the service was down or experiencing issues over the weekend, that might explain the intermittent problems.

Kindly refer below link: authentication-failed

Thank You.

Answer 2

Here is Azure's support on why it was a server error last Friday, June 20, 2025:
“The failure was caused by an error in the logic that selects the audio transcription model, specifically in reconciling customer input with the system default. When a customer specifies the model name in the request payload, the service returns an internal server error. The request is processed normally if no model name is provided, but the model field is required in some recent API versions. This logic resides in the Completions API endpoint service. The defect passed our validation gate and reached production because we lacked test cases for API versions that support specifying a model name in the payload”.