![]() ![]() The New-AzSSMultiBatchRequest is the function that leverages most of the other functions to get its work done. New-AzSSMultiBatchRequest: creates a transcription request for every blob in a container, waits for each to complete, and saves result files to a directory.New-AzStorageSASTokenAllBlobs: creates a SAS token for every blob item in a container on an Azure Storage Account, required for the Speech Service to access the blob.Remove-AzSSBatchRequest: removes a completed transcription request, whether is was successful or not.New-AzSSBatchRequest: creates a new transcription request, and returns the URI of the request.Get-AzSSBatchResults: retrieves the JSON result files from a successful transcription.Get-AzSSBatchStatus: retrieves the current status of a particular transcript request ID.There are six functions that compose what I needed. ![]() Why not just write the whole thing with PowerShell functions? And that is exactly what I did. But I use PowerShell a lot! And PowerShell can talk to a Rest API using Invoke-WebRequest or Invoke-RestMethod. The batch service examples in particular are only available in C#. If you want to use the Speech Service SDK, there are examples on GitHub. That is more like it! Let’s do that.Įxcept, the Rest API is just an API, not a GUI or a menu. The batch process will transcribe the file, and then you can retrieve the results using another call to the API. There is also a Rest API that supports batch processing, allowing you to send a request to the transcription endpoint with a file stored in Azure Storage. The point of that is to integrate speech to text in your applications. The service itself can be used in concert with the Speech SDK to convert snippets about 15 seconds long. That is not exactly what the Speech Service does. Then it would dutifully transcribe all the files and dump out the transcriptions in some file repository. When I decided to try and use the Azure Speech Service, I initially assumed that I would be able to upload the files to Azure Storage and point the service at all my MP3s. What if I could transcribe the audio to text, and then search through the text to find all the times we talked about Derrick and Miranda, how we’re all doomed, or smiling poop? The Azure Speech to Text API can be used to convert speech to text of audio files. There’s no way I could listen to the entirety of the episodes, and so I started thinking. At an average of 35 minutes, that’s roughly 57 hours of combined audio. In preparation for the 100th episode, I thought it might be nice to look over past episodes and find some common themes, running gags, and anything else that caught my eye. As I write this, we are getting ready to record episode 98. The 100th episode of Buffer Overflow – a weekly tech news podcast I host – is steadily approaching.
0 Comments
Leave a Reply. |