crispengari/speech-to-text-python-ibm_watson

This is a simple Artificial Intelligence Application that converts audios to speech using `ibm_watson`.

Speech To Text (stt)

This is a simple Artificial Intelligence Application that converts audios to speech using ibm_watson.

Application capabilities

This app is capable of:

• reading an audio from audios folder
• converts the audio to speech using ibm_watson SpeechToTextV1()
• write the converted speech to an external file speech.txt in the files folder

Getting started

Installation

####### First you need to install ibm_watson

$pip install ibm_watson

Second you need to install pip install PyJWT==1.7.1

$pip install PyJWT==1.7.1

Then you are ready to go

Getting an API key and service URL

To get the service URL go to IBM WATSON

• Create an account or login if you are a member
• Go to service
• Go to AI
• Look for Speech To Text and click
• Create a new project
• Hunt for API keys in the docs

Importing packages

from ibm_watson import SpeechToTextV1, ApiException
from ibm_cloud_sdk_core.authenticators import IAMAuthenticator
import json

Keys Variables

url = "API_KEY"
api_key = "URL"

Setting the authentication

try:
    auth = IAMAuthenticator(api_key)
    stt = SpeechToTextV1(authenticator=auth)
    stt.set_service_url(url)
except ApiException as e:
    print(e)

Converting audio to speech

with open("audios/long.mp3", "rb") as audio:
    res = stt.recognize(audio=audio, content_type="audio/mp3", model="en-AU_NarrowbandModel", continuous=True).get_result()

Write all the speech in a text file

sentences = res["results"]
sentence_list = []
for sentence in sentences:
    # adding a sentence with confidence that is greater than 50%
    sentence_list.append(str(sentence["alternatives"][0]["transcript"]).strip() if sentence["alternatives"][0]["confidence"] > 0.5 else "")

# print(json.dumps(sentence_list, indent=2))

with open("files/speech.txt", "w") as writter:
    for line in sentence_list:
        if line == "%HESITATION":
            writter.write(",")
        else:
            writter.write(line+" ")

print("DONE")

All the code in one file main.py

# importing packages
from ibm_watson import SpeechToTextV1, ApiException
from ibm_cloud_sdk_core.authenticators import IAMAuthenticator
import json

# service credentials
url = "API_KEY"
api_key = "URL"

# Setting the authentication
try:
    auth = IAMAuthenticator(api_key)
    stt = SpeechToTextV1(authenticator=auth)
    stt.set_service_url(url)
except ApiException as e:
    print(e)

# converting audio to speech
with open("audios/long.mp3", "rb") as audio:
    res = stt.recognize(audio=audio, content_type="audio/mp3", model="en-AU_NarrowbandModel", continuous=True).get_result()


""""
* We are getting a python list of number of results
* We want to loop through them and create sentences
"""

sentences = res["results"]
sentence_list = []
for sentence in sentences:
    # adding a sentence with confidence that is greater than 50%
    sentence_list.append(str(sentence["alternatives"][0]["transcript"]).strip() if sentence["alternatives"][0]["confidence"] > 0.5 else "")

# print(json.dumps(sentence_list, indent=2))

with open("files/speech.txt", "w") as writter:
    for line in sentence_list:
        if line == "%HESITATION":
            writter.write(",")
        else:
            writter.write(line+" ")

print("DONE")

Changes

There are a list of models and you can change the code based on what you want to achive

• Modes URL HERE
• Speech To Text Docs HERE

Why this simple Application.

This program was built for practical purposes

Credits:

• None