I am using OpenAI whisper model to generate transcription for call recordings.
I am have verbose_json as response_format.
But intermittently, the model is generating ‘…’ as response sometimes.
We have identified this has happened only for non-english audio up until now.
Here is the sample image on what it generated
Also here the code snippet, if it helps.
with open(audio_file_path, "rb") as audio_file:
return openai.Audio.transcribe("whisper-1", audio_file, response_format="verbose_json")