Batch transcription with Microsoft Azure (REST API)

Question

I want transcribe longer audio files (at least 5 minutes) using REST APIs from Microsoft. There are a lot of different products and names, e.g. Speech service API or Bing Speech API. None of the REST APIs I tried so far supports transcribing longer audio files.

The documentation states there is a REST API exactly for this case: https://learn.microsoft.com/en-us/azure/cognitive-services/speech-service/batch-transcription

What is the endpoint for this service?

Tel

Answer 1

There is a sample available on GitHub here: https://github.com/PanosPeriorellis/Speech_Service-BatchTranscriptionAPI

The endpoint is CRIS's endpoint, as in this code:

private const string HostName = "cris.ai";
// ...
var client = CrisClient.CreateApiV2Client(SubscriptionKey, HostName, Port);

Then I found on the documentation that the API is exposed on Swagger (link visible here), so it's easier to explore the methods available (switch from 2.0beta to 2.0 on top):

• For West Europe: https://westeurope.cris.ai/swagger/ui/index
• For West US : https://westus.cris.ai/swagger/ui/index

So to create a new transcription, the path is: /api/speechtotext/v2.0/transcriptions, called with the POST method, so the full endpoint is:

• For West Europe: https://westeurope.cris.ai/api/speechtotext/v2.0/transcriptions
• For West US : https://westus.cris.ai/api/speechtotext/v2.0/transcriptions

Please note that the level of your subscription key needed to use the transcription must be a Standard level pricing S0, not Free one.

Baillargeon

Thank you for the answer. This doesn't help yet. Which URL path etc. is this? I tried https://westeurope.cris.ai/api/speechtotext/v2.0/transcriptions with a subscription key from https://westeurope.cris.ai/Subscriptions and I get error 400 "Only "Standard" subscriptions for the region of the called service are valid." – Tel

So it looks like you have to get an API key of "standard" pricing. Your current key was a free one? – Baillargeon

yes. the problem was that I needed a S0 key, see github.com/MicrosoftDocs/azure-docs/issues/… can you add the details about the subscription key to your answer? then I will accept it as the right answer – Tel

@Tel were you able to transcribe multiple files with a single API call? If so can you explain how? I am only able to successfully transcribe a single file with this API and can't find multi-file documentation. – Iceblink

@Iceblink I only transcribed files one by one with multiple calls. – Tel

I think I am very late in responding to this question. But there is a way you can do batch transcription . Please check this learn.microsoft.com/en-us/azure/cognitive-services/… or more details – Gangling

@HarmandeepSinghKalsi at the time the batch transcription didn't work. I haven't tested this lately so it may be more up to date. – Iceblink

Yeah , now they have released v3.0 which has batch transcription API wherein you can just give the container path and all the files will be considered for the transcription – Gangling