Azure-docs: batch transcription in python

[kiranmahto]

Can you give some samples on how to use batch transcription with python.

---

Document Details

⚠ Do not edit this section. It is required for docs.microsoft.com ➟ GitHub issue linking.

• ID: 73fb7e76-fb80-c420-2efe-e3ca3a27b749
• Version Independent ID: 07a42557-deb3-e4b3-3f14-f26a27be49c2
• Content: How to use Batch Transcription - Speech Services
• Content Source: articles/cognitive-services/Speech-Service/batch-transcription.md
• Service: cognitive-services
• GitHub Login: @PanosPeriorellis
• Microsoft Alias: panosper

[CHEEKATLAPRADEEP-MSFT]

@kiranmahto Thank you for your interest in Azure products and services. This is being assigned to the content author to have a look and update as appropriate.

[kishan19]

This will be very helpful

[deivit78]

Whenever I am told to use azure with something other than .NET it doesn't work or causes a lot of problems.
I think that when you publish an API, you should have examples and documentation.
I don't understand how a company like Microsoft leaves aside the rest of the world that has to use languages like python, even more in the field of data science.

[PanosPeriorellis]

There are details of a Python sample being uploaded today. We are also uploading related documentation on how to automatically generate client libraries using our Swagger docs.

[deivit78]

Testing the python script:
Connected to pydev debugger (build 191.6605.12)
Traceback (most recent call last):
File "/home/dave/programas/pycharm-2019.1.1/helpers/pydev/pydevd.py", line 1741, in
main()
File "/home/dave/programas/pycharm-2019.1.1/helpers/pydev/pydevd.py", line 1735, in main
globals = debugger.run(setup['file'], None, None, is_module)
File "/home/dave/programas/pycharm-2019.1.1/helpers/pydev/pydevd.py", line 1135, in run
pydev_imports.execfile(file, globals, locals) # execute the script
File "/home/dave/programas/pycharm-2019.1.1/helpers/pydev/_pydev_imps/_pydev_execfile.py", line 18, in execfile
exec(compile(contents+"\n", file, 'exec'), glob, loc)
File "/home/dave/PycharmProjects/azure/main.py", line 8, in
import swagger_client as cris_client
File "/home/dave/PycharmProjects/googleTest/azure/lib/python3.6/site-packages/swagger_client/__init__.py", line 19, in
from swagger_client.api.custom_speech_accuracy_tests_api import CustomSpeechAccuracyTestsApi
File "/home/dave/PycharmProjects/googleTest/azure/lib/python3.6/site-packages/swagger_client/api/__init__.py", line 6, in
from swagger_client.api.custom_speech_accuracy_tests_api import CustomSpeechAccuracyTestsApi
File "/home/dave/PycharmProjects/googleTest/azure/lib/python3.6/site-packages/swagger_client/api/custom_speech_accuracy_tests_api.py", line 21, in
from swagger_client.api_client import ApiClient
File "/home/dave/PycharmProjects/googleTest/azure/lib/python3.6/site-packages/swagger_client/api_client.py", line 27, in
import swagger_client.models
File "/home/dave/PycharmProjects/googleTest/azure/lib/python3.6/site-packages/swagger_client/models/__init__.py", line 23, in
from swagger_client.models.endpoint import Endpoint
File "/home/dave/PycharmProjects/googleTest/azure/lib/python3.6/site-packages/swagger_client/models/endpoint.py", line 19, in
from swagger_client.models.model import Model # noqa: F401,E501
File "/home/dave/PycharmProjects/googleTest/azure/lib/python3.6/site-packages/swagger_client/models/model.py", line 20, in
from swagger_client.models.model import Model # noqa: F401,E501
ImportError: cannot import name 'Model'

[deivit78]

I don't know if the problem is:
from swagger_client.models.model import Model # noqa: F401,E501
class Model(object):

It seems ambiguous.

[JamesEarle]

@deivit78 I also experienced the same error listed above with the Python sample. There is clearly some abstraction done because the JSON response to the API is fairly complex, and actually requires multiple calls, but you can simply use Python requests to write request like below.

POST https://<your-speech-region>.cris.ai/api/speechtotext/v2.0/Transcriptions

Headers
{
"Ocp-Apim-Subscription-Key": "your-subscription-key"
"Content-Type": "application/json"
}

Body
{
  "recordingsUrl": "a generated SAS URL to your audio file in an Azure Storage Account ",
  "models": [],
  "locale": "en-US",
  "name": "Transcription using locale en-US",
  "description": "An optional description of the transcription.",
  "properties": {
    "ProfanityFilterMode": "Masked",
    "PunctuationMode": "DictatedAndAutomatic"
  }
}

You can add {"id": "<your-custom-language-model-id>"} into the models array in the body of the payload if you want to use a custom model, otherwise it will use the most up to date unified speech model.

The response body to that request will contain a Location which is a URL you can do a GET against and that is where your transcripts will be, but the Retry-After in the response is the number of seconds you must wait before checking the location (according to the docs) as the transcription can take time.

[codinguncut]

I am incredibly pissed of with microsofts documentation and usability on this.

• I had the same issue re: invalid swagger_client python code
• why does the normal azure api only support wav? I can see no possible reason for this
• why does microsoft need so many domains (cris.ai, videoindexer.ai, etc.) which are just making life harder re: cookies, etc.
• then I got HTTP response body: {"code":"InvalidProductId","message":"The subscription SKU \"CognitiveServices.S0\" is not supported in this service instance."}
• then I upgraded my free azure account to a pay-as-you-go
• it took ages for azure to realize that my subscription had been updated
• Now I'm getting HTTP response body: {"code":"Unauthorized","message":"Authentication is required to access the resource."}

Dear Microsoft, the year is 2019 and you are desperately trying to compete with AWS, everyone else just provides an sdk that works. I have never, ever had this many unnecessary problems with AWS. Please provide documentation and sane ways of interacting with your services. It's honestly not that hard. Absolutely shocking!

[theelk801]

I'm also very frustrated by this, I'd like to be able to run speech analysis on audio files but the documentation is nigh-incomprehensible. Honestly I'd use a different service but this project requires me to use Azure so I don't really have a choice.

[chlandsi]

Sorry that you are experiencing problems. We have updated the sample code and instructions with workarounds for a swagger bug (import Model failed) and a more detailed instructions. For the authentication problem, make sure that the region given for the endpoint in your code and the region you used to download the swagger client are the same. If you continue to have problems with the sample, consider opening a issue in the samples repo.

[tchristiani]

Please add the label speech-service/subsvc to this issue.
Thank you.

[tchristiani]

Sample has been updated to address this issue.

please-close

[RyanCantab]

Agreed that this is ridiculous, why has it taken me over half a day to work out how to get this service to do what I want? There are some code snippets that show how to make it work for mic input under 15 seconds long. That's it. MS have forgotten to document their own product. What exactly was the plan here? Bizarre and annoying.