Blob Trigger function in azure for speech-to-text in python

Question

My code looks like below and it is setupped in azure in such a way that the trigger is happening properly but in result I'm getting some buffer or encoded values to be check:

1. output result is not genrating a plain text as a transcription result we are getting some audio buffer or an encoded results that need to be fixed 2.sending the audio files they are in .wav format itself hence their is no format issue i thing

import logging
import azure.functions as func
import azure.cognitiveservices.speech as speechsdk
from azure.storage.blob import BlobServiceClient, ContentSettings

# Azure Speech Service configuration

speech_key = "speech-key"
service_region = "eastus"

# Configure Speech Recognizer

speech_config = speechsdk.SpeechConfig(subscription=speech_key, region=service_region)

def main(blob: func.InputStream, outputBlob: func.Out\[str\]):
logging.info(f"Python blob trigger function processed blob\\n"
f"Name: {blob.name}\\n"
f"Blob Size: {blob.length} bytes")

    # Get the audio data from the blob
    audio_data = blob.readall()
    
    # Create an audio stream from the audio data
    audio_stream = speechsdk.audio.AudioDataStream(audio_data)
    
    # Create a speech recognizer
    speech_recognizer = speechsdk.SpeechRecognizer(speech_config=speech_config, audio_config=audio_stream)
    
    # Perform speech recognition
    result = speech_recognizer.recognize_once()
    
    # Get the transcript
    transcript = result.text if result.reason == speechsdk.ResultReason.RecognizedSpeech else ""
    
    # Save transcript as a text file
    output_file = "output.txt"
    with open(output_file, "w") as file:
        file.write(transcript)
    
    # Save transcript to a new blob
    connection_string = "DefaultEndpointsProtocol=https;AccountName=scribeemrnewtesting8a49;AccountKey=RvyepyfHuKQjIrbPLSnx36zNB65l64kUurdSFi903DhXdz+pzbFTgpuNk6yviESmPlIVDFqMkjxU+AStRPdFAA==;EndpointSuffix=core.windows.net"
    output_container_name = "text-output"
    blob_name = blob.name.replace(".wav", ".txt")
    
    blob_service_client = BlobServiceClient.from_connection_string(connection_string)
    output_container_client = blob_service_client.get_container_client(output_container_name)
    
    # Upload transcript to output blob container/
    output_blob_client = output_container_client.get_blob_client(blob_name)
    with open(output_file, "rb") as file:
        output_blob_client.upload_blob(file, overwrite=True, content_settings=ContentSettings(content_type='text/plain'))
    
    blob.delete()
    
    logging.info(f"Transcription completed and saved to the out-blob container: {transcript}")
    outputBlob.set(transcript)

getting output as: RIFF�� WAVEfmt �> } data�� �������� �� ! �� �� �� % ���� "

type here

expected : A plain text which has transcription as per the above logic

Answer 1

Firstly, I created a container for input blob and didn't uploaded any blob to that as below,

And, I created another container for output blob and didn't upload any blob to that as below,

I made some changes with your code and got the expected text output with input .wav blob to .txt blob and also the transcription of output.txt.

Code:

import logging
import azure.functions as func
import azure.cognitiveservices.speech as speechsdk
from azure.storage.blob import BlobServiceClient, ContentSettings
import tempfile

speech_key = "<key>"
service_region = "<region>"

speech_config = speechsdk.SpeechConfig(subscription=speech_key, region=service_region)

def main(blob: func.InputStream, outputBlob: func.Out[str]):
    logging.info(f"Python blob trigger function processed blob\n"
                 f"Name: {blob.name}\n"
                 f"Blob Size: {blob.length} bytes")

    try:
        audio_data = blob.read()

        with tempfile.NamedTemporaryFile(delete=False) as temp_file:
            temp_file.write(audio_data)
            temp_file.seek(0)
            audio_config = speechsdk.audio.AudioConfig(filename=temp_file.name)
            speech_recognizer = speechsdk.SpeechRecognizer(speech_config=speech_config, audio_config=audio_config)
            result = speech_recognizer.recognize_once()

        transcript = result.text if result.reason == speechsdk.ResultReason.RecognizedSpeech else ""
        output_file = "output.txt"
        with open(output_file, "w") as file:
            file.write(transcript)

        connection_string = "<storage-connec-string>"
        output_container_name = "text-output"
        blob_name = blob.name.replace(".wav", ".txt")

        blob_service_client = BlobServiceClient.from_connection_string(connection_string)
        output_container_client = blob_service_client.get_container_client(output_container_name)

        output_blob_client = output_container_client.get_blob_client(blob_name)
        with open(output_file, "rb") as file:
            output_blob_client.upload_blob(file, overwrite=True, content_settings=ContentSettings(content_type='text/plain'))

        logging.info(f"Transcription completed and saved to the out-blob container: {transcript}")
        outputBlob.set(transcript)

    except Exception as ex:
        logging.error(f"Exception occurred during speech recognition: {ex}")
        raise ex

function.json:

{
  "scriptFile": "__init__.py",
  "bindings": [
    {
      "name": "blob",
      "type": "blobTrigger",
      "direction": "in",
      "path": "audio-input/sample.wav",
      "connection": "AzureWebJobsStorage"
    },
    {
      "name": "outputBlob",
      "type": "blob",
      "direction": "out",
      "path": "text-output/sample.txt",
      "connection": "AzureWebJobsStorage"
    }
  ]
}

Replace the input and output container names and blob names in above function.json file

local.settings.json:

{
  "IsEncrypted": false,
  "Values": {
    "FUNCTIONS_WORKER_RUNTIME": "python",
    "AzureWebJobsStorage": "<storage-account-conne-string>",
    "AzureWebJobsFeatureFlags": "EnableWorkerIndexing"
}
}

requirement.txt:

azure-functions
azure-cognitiveservices-speech==1.21.0
azure-storage-blob

Output:

The blob trigger function runs successfully and there is no text generated at output.txt file as below,

Now, I uploaded a sample.wav file to my input container as below,

Then, the output.txt file got generated with some text of the sample.wav blob as below,

And a sample.txt file got created in my text-output container as below,

Then, I downloaded the output blob to check the text output as below,

The text got generated in output blob as below,