What we're building
This is the fundamental loop:
- The assistant hears what you have to say.
- That speech becomes text as a result.
- It determines what you desire.
- It responds to you aloud.
That's all. Basic cycle. It reacts when you speak.
Let's keep it brief: start by stating the time, look up information on Wikipedia, and then crack a joke. Additional functions like setting reminders or checking the weather can be added later.
Tools you'll need
- Python 3 is installed on your computer.
- A mic (your laptop mic is fine)
- Internet (for the speech recognition part)
- These Python libraries:
-
- SpeechRecognition
-
- pyttsx3
-
- pyaudio
-
- wikipedia
-
- pyjokes
Run this to install everything:
Pip install SpeechRecognition pyttsx3 pyaudio wikipedia pyjokes
Sometimes pyaudio is a pain to install. If it fails, Try Searching for "install pyaudio wheel” for your system; it’s a common fix.
Step 1: Make it talk
Before we teach it to listen, let's get the speaking part working.
import pyttsx3
engine = pyttsx3.init()
def speak(text):
print(f"Assistant: {text}")
engine.say(text)
engine.runAndWait()
speak("Hello, I’m ready.")
Run this. If your computer talks back, you’re good.
Step 2: Make it listen
Now let’s capture your voice and turn it into text.
import speech_recognition as sr
def listen():
r = sr.Recognizer()
with sr.Microphone() as source:
print("Listening...")
audio = r.listen(source)
try:
text = r.recognize_google(audio)
print(f"You said: {text}")
return text.lower()
except sr.UnknownValueError:
speak("Sorry, I didn’t get that.")
return ""
except sr.RequestError:
speak("Speech service is down.")
return ""
Now it listens to you and prints what you said. If it can’t figure it out, it just says sorry instead of breaking.
Step 3: Add commands
Okay, let’s teach it what to do when it hears you.
import datetime
import wikipedia
import pyjokes
def handle_command(command):
if "time" in command:
now = datetime.datetime.now().strftime("%I:%M %p")
speak(f"The time is {now}")
return True
if "wikipedia" in command:
topic = command.replace("wikipedia", "").strip()
if not topic:
speak("Tell me what to search on Wikipedia.")
return True
speak("Searching Wikipedia...")
try:
summary = wikipedia.summary(topic, sentences=2)
speak(summary)
except:
speak("I couldn’t find that topic.")
return True
if "joke" in command:
joke = pyjokes.get_joke()
speak(joke)
return True
if "exit" in command or "quit" in command:
speak("Goodbye.")
return False
speak("I don’t know how to do that yet.")
return True
Then glue everything together:
if __name__ == "__main__":
speak("Hello, how can I help?")
running = True
while running:
query = listen()
if query:
running = handle_command(query)
Now test it. Say:
- “time” → it tells you the time.
-
- “Wikipedia Python” → it reads a short summary.
-
- “tell me a joke” → you’ll get something silly.
-
- “exit” → it stops.
-
Step 4: Improve its usability
It's rough, but it works for now. Some immediate enhancements:
To account for background noise, use:
(source, duration=0.5) r.adjust_for_ambient_noise
Answers should be brief. Prolonged speeches are dull.
Address errors with cordial responses.
These minor details count. They let the assistant feel more like an assistant and less like a robot.
Step 5: Add more features
Once the basics work, adding new tricks is fun.
Open YouTube:
import webbrowser
if "open youtube" in command:
webbrowser.open("https://youtube.com")
speak("Opening YouTube.")
return True
Weather reports: use the OpenWeatherMap API.
Reminders: use the schedule library.
To-do list: save and read from a text file.
Each new feature is just another “if” block.
Step 6: Personality matters
Nobody likes a monotone assistant. Change the voice and pace.
voices = engine.getProperty('voices')
engine.setProperty('voice', voices[1].id) # Switch voice
engine.setProperty('rate', 175) # Adjust speed
You can also make it greet you differently in the morning, afternoon, and evening. Small touches make it fun.
Step 7: Organize your code
If you keep everything in one file, it’ll get messy fast. Break it into files:
- speech_io.py → listening and speaking
- commands.py → what it can do
- main.py → runs the loop
Cleaner, easier to grow.
Step 8: What about “Hey Assistant”?
That’s called a wake word. It’s possible, but it takes more resources. For now, just press Enter to start listening or use a hotkey. Later, you can explore proper wake word engines.
Step 9: Add a simple GUI
With tkinter, you can make a tiny window with a mic button and a box that shows what was said. Not required, but handy if you want to demo it to friends.
Step 10: Expect errors
Sometimes the mic doesn’t pick up. Sometimes the speech API is down. Don’t sweat it—just handle the errors gracefully. Say “I can’t connect” instead of freezing.
Where to take this next
You now have a working voice assistant. From here, you can:
- Pull in daily news headlines.
- Add translations.
- Send quick emails.
- Play music.
- Take voice notes.
-
Each little project adds new skills.
And if you find yourself thinking “this is fun, I want to really learn AI,” that’s when a structured course comes in. The AI using Python course in Noida by Uncodemy is a good fit here because it moves beyond hobby projects into serious AI development—natural language processing, data handling, and building smarter apps.
Final thoughts
You can actually build your own voice assistant. You don't require a lot of code or expensive equipment. Only a microphone, a few libraries, and Python.
Start small with jokes, Wikipedia, and time. Next, enlarge. Consider adding a GUI, additional functionality, and personality. You'll have an assistant who truly feels helpful before you know it.
And don't just keep experimenting aimlessly if you want to develop this curiosity into a legitimate ability. Learn the fundamentals using an organized course, such as Uncodemy's AI using Python course. That serves as a link between "weekend projects" and "real-world AI skills."
So take out your laptop, type a few words, and greet your new helper. The speed at which it comes to life will amaze you.
