Diarization does not work for Greek with OpenAI Whisper model

Nik Uspenskiy 0 Reputation points
2023-12-04T15:28:35.2233333+00:00

Hello. I am trying to separate two speakers in audio in Greek with Azure Batch Transcription service. The audio is transcribed correctly but the diarization feature does not work for Greek while I had no problem with Romanian.

Please find below the code for Batch Transcription.

curl -v -X POST -H "Ocp-Apim-Subscription-Key: YourSubscriptionKey" -H "Content-Type: application/json" -d '{   "contentUrls": ["http://www.skipperguru.ru/Files/AIVOTAR/GR/SCRIPT%201.3.mp3"],   "locale": "el-GR",   "displayName": "My Transcription",   "model": {     "self": "https://eastuss0.cognitiveservices.azure.com/speechtotext/v3.2-preview.1/models/base/e830341e-8f47-4e0a-b64c-3f66167b751c"   },   "properties": {     "diarizationEnabled": true,     "languageIdentification": {       "candidateLocales": [         "en-US", "el-GR"       ],     }   }, }'  "https://YourServiceRegion.cognitiveservices.azure.com/speechtotext/v3.2-preview.1/transcriptions"

Both audio files in Greek and Romanian are in MP3 format. And in both cases, OpenAI whisper V2 model is used.

Please help to identify why the diarization feature does not work for Greek.

Azure AI services
Azure AI services
A group of Azure services, SDKs, and APIs designed to make apps more intelligent, engaging, and discoverable.
2,606 questions
{count} votes