Voice Assistant with Wake Word in Python

In this article, you will learn how to build an intelligent voice assistant in Python that reacts to a wake word. The assistant will not process any sound unless it hears the specific wake word you designate. Here, we'll name our assistant "Jake" and use "Hey Jake" as the wake word. The assistant will activate upon hearing the wake word and perform a set of predefined tasks based on your commands.

Demo Preview

Before diving into the implementation, let's see what we aim to achieve. Running the script initializes a graphical user interface (GUI), showcasing a bot icon. When the bot is inactive, the icon is black. It turns red upon hearing the wake word ("Hey Jake"), indicating it's ready to process commands. Commands such as "What is neural nine?" will be processed and answered, and the bot reverts to an inactive state, turning the icon back to black.

Project Dependencies

To build this voice assistant, we need three external Python modules:

speech_recognition for converting speech to text.
neuralintents for the core assistant functionality.
pyttsx3 for turning text responses into speech output.

Install these modules using pip:

pip install speechrecognition
pip install neuralintents
pip install pyttsx3

Setting Up the Assistant Class

Create a new Python class named Assistant to encapsulate the entire voice assistant logic. Import the necessary libraries and set up the recognizer, speaker, and assistant model.

import sys
import threading
import tkinter as tk
import speech_recognition as sr
import pyttsx3 as tts
from neuralintents import GenericAssistant

class Assistant:
    def __init__(self):
        self.recognizer = sr.Recognizer()
        self.speaker = tts.init()
        self.speaker.setProperty('rate', 150)
        
        self.assistant = GenericAssistant('intents.json', intent_methods=('file': self.create_file))
        self.assistant.train_model()
        
        self.root = tk.Tk()
        self.label = tk.Label(self.root, text='?', font=('Arial', 120))
        self.label.pack()
        
        threading.Thread(target=self.run_assistant).start()
        self.root.mainloop()

Creating the Intents File

Create a JSON file named intents.json to determine the bot's capabilities. This file will include tags, patterns, and responses for different commands.

(
    "intents": [
        {
            "tag": "greeting",
            "patterns": ["Hi", "Hey", "What's up?", "Hello"],
            "responses": ["Hello!", "How can I assist you?"]
        ),
        (
            "tag": "neural_9",
            "patterns": ["What is neural nine?", "Tell me about neural nine?"],
            "responses": ["Neural 9 is a YouTube channel focused on programming."]
        ),
        (
            "tag": "name",
            "patterns": ["What's your name?", "Who are you?"],
            "responses": ["My name is Jake."]
        ),
        (
            "tag": "file",
            "patterns": ["Create a file"],
            "responses": []
        )
    ]
}

Running the Assistant

The run_assistant function is responsible for continuously listening to audio input, checking for the wake word, and processing subsequent commands.

    def run_assistant(self):
        while True:
            try:
                with sr.Microphone() as mic:
                    self.recognizer.adjust_for_ambient_noise(mic, duration=0.2)
                    audio = self.recognizer.listen(mic)
                    text = self.recognizer.recognize_google(audio).lower()
                    
                    if "hey jake" in text:
                        self.label.config(fg='red')
                        
                        audio = self.recognizer.listen(mic)
                        text = self.recognizer.recognize_google(audio).lower()
                        
                        if text == "stop":
                            self.speaker.say("Bye")
                            self.speaker.runAndWait()
                            self.speaker.stop()
                            self.root.destroy()
                            sys.exit()
                        
                        if text:
                            response = self.assistant.request(text)
                            if response:
                                self.speaker.say(response)
                                self.speaker.runAndWait()
                        
                        self.label.config(fg='black')
                
            except Exception as e:
                print(e)
                self.label.config(fg='black')
                continue

Creating a File

The create_file method demonstrates how to connect a command to a Python function. In this case, the function creates a text file.

    def create_file(self):
        with open("some_file.txt", "w") as f:
            f.write("Hello World")

Running the Script

Create an instance of the Assistant class to run the script.

if __name__ == "__main__":
    Assistant()

Running this script will activate your voice assistant, Jake. It will listen for the wake word, process commands, and perform actions based on your inputs.

Keyword

Voice Assistant
Wake Word
Python
Speech Recognition
Neuralintents
Pyttsx3
GUI

FAQ

1. What modules are required to build this voice assistant?

You need speech_recognition, neuralintents, and pyttsx3.

2. How can I install the required modules?

Use pip to install the modules:

pip install speechrecognition
pip install neuralintents
pip install pyttsx3

3. What is the purpose of the intents file?

The intents file specifies the different commands the assistant can handle, including sample phrases (patterns) and responses.

4. How does the assistant activate?

The assistant activates upon hearing the designated wake word, "Hey Jake", and then processes the following command.

5. How can I customize the assistant’s responses?

Modify the responses field in the intents.json file to customize the assistant’s replies.

6. Can I add more functionalities to the assistant?

Yes, you can add more functionalities by defining new tags and corresponding functions in the intents.json file and your script.