Unable to get word level timestamp from AzureOpenAI client, whisper-1

API Bugs
,
1

I am using AzureOpenai client, with that i am not able to get word level granularity, but with openai client it gives the data properly!

Here is a short code snippet


from openai import OpenAI, AzureOpenAI

openai_client = OpenAI()
azure_client = AzureOpenAI(
                api_key=api_key,
                api_version="2024-02-15-preview",
                azure_endpoint=azure_endpoint
            )

audio_file = open('test.m4a', "rb")
azure_transcript = azure_client.audio.transcriptions.create(
  file=audio_file,
  model="whisper-1",
  response_format="verbose_json",
  timestamp_granularities=['word']
)
openai_transcript = openai_client.audio.transcriptions.create(
  file=audio_file,
  model="whisper-1",
  response_format="verbose_json",
  timestamp_granularities=['word']
)

print(openai_transcript.words) ==> Gives correct results
print(azure_transcript.words) ==> returns None, azure_transcript.segments has the data, but they are not based on word level!