Custom endpoint available for ConversationTranscriber?

Marc Sumner 0

I have followed the instructions to train a new speech model and created a custom endpoint for it in Speech Studio. I have seen examples for using the custom endpoint for SpeechRecognizer; however, when I try to connect to the endpoint using a ConversationTranscriber it seems to simply use the default endpoint, not my custom one. I also do not see any error messages that my endpoint is not available.

Question is: is it possible to train a model for a ConversationTranscriber and use it at a custom endpoint or is this not supported?

dupammi 8,030 Reputation points Microsoft Vendor

2024-07-19T09:55:07.1366667+00:00

Hi @Marc Sumner

Thank you for using the Microsoft Q&A forum.

It is not currently possible to use a custom endpoint with the ConversationTranscriber in Azure Speech service. The ConversationTranscriber uses the default endpoint provided by Azure and does not support specifying a custom endpoint for your trained model. If you require the use of a custom model, you may need to handle speaker diarization and transcription manually using the SpeechRecognizer with your custom endpoint and implement additional logic to manage conversation context and speaker turns.

For more details please refer - implement-diarization-from-file-with-conversation-transcription

and Quickstart: Create real-time diarization

Share via

Custom endpoint available for ConversationTranscriber?