Title
Turn Audio Recordings into Perfect Meeting Minutes in 1 Minute! Build Your Own 'AI Secretary' ๐๏ธ๐
Introduction
Hello, creators! Welcome to a2set's AI Tutorial.
Have you ever recorded an important meeting, Zoom call, or college lecture? While recording gives you peace of mind, listening to a 1-hour audio file all over again to summarize the core points and type them out is absolute torture.
Today, using Google Colab and OpenAI's powerful AI models, we will build your very own "AI Secretary." It will automatically transcribe uploaded audio files (Whisper) and organize them into a clean meeting minute format, complete with core topics and Action Items (GPT-4o).
This single script will 10x your work and study productivity. Keep your browser open and follow right along!
๐จ Mandatory Check Before We Start (OpenAI API Billing)
To borrow OpenAI's brain for this tutorial, you need an API key and a pre-loaded balance of at least $5 in your account. Transforming a 1-hour recording into perfect meeting minutes costs about the price of a sip of coffee (approx. $0.50). Think of it as the monthly salary for your lifelong AI secretary!
Step 1: Google Colab Setup & OpenAI API Prep
Let's prepare our free digital workspace (Google Colab) to run Python code.
Go to Google Colab and click the blue [New Notebook] button. Elegantly rename the file to AI_Meeting_Minutes.ipynb.
Go to the OpenAI API Billing Page, charge a minimum of $5, then generate and copy a new API key (sk-...) from the API keys menu.
In the first cell of Colab, paste the code to install libraries and set your API key, then hit the โถ๏ธ (Play button).
# Install OpenAI library
!pip install openai
import os
from openai import OpenAI
# Insert your secret API key
os.environ["OPENAI_API_KEY"]
# Install OpenAI library
!pip install openai
import os
from openai import OpenAI
# Insert your secret API key
os.environ["OPENAI_API_KEY"]
# Install OpenAI library
!pip install openai
import os
from openai import OpenAI
# Insert your secret API key
os.environ["OPENAI_API_KEY"]
Step 2: Upload Audio File (MP3)
Now it's time to bring your meeting or lecture recording file (.mp3, .m4a, etc.) from your smartphone to Colab.
Run the script below in a new code cell, and a file selection window will appear.
from google.colab import files
print("๐๏ธ Please upload your audio file (mp3, m4a, wav)...")
# Upload file from your PC
uploaded = files.upload()
# Get the exact file name
audio_file_name = list(uploaded.keys())[0]
from google.colab import files
print("๐๏ธ Please upload your audio file (mp3, m4a, wav)...")
# Upload file from your PC
uploaded = files.upload()
# Get the exact file name
audio_file_name = list(uploaded.keys())[0]
from google.colab import files
print("๐๏ธ Please upload your audio file (mp3, m4a, wav)...")
# Upload file from your PC
uploaded = files.upload()
# Get the exact file name
audio_file_name = list(uploaded.keys())[0]
After running, click the [Choose Files] button to upload your recording.
(Due to the Whisper model's limitations, the file size must be under 25MB. If it's too long, please compress or split the file before uploading.)
Step 3: Transcribe Audio to Text with Whisper AI (STT)
It's time to use the incredible Whisper model. As the best speech recognition AI available today, it perfectly understands and converts speech to text, even with slightly muffled pronunciation or background noise.
We've added code to safely open the file and prevent errors.
Hit the play button! Depending on the file size, in just a few seconds to a minute, the entire conversation will be extracted as perfect text!
Step 4: Enter the Executive AI Secretary! (Structuring Perfect Meeting Minutes)
Now, we will hand over this long, raw transcript to GPT-4o and command it to summarize it into a clean, professional 'Meeting Minutes' document.
Open a new code cell and run the code below. The prompt instructs the AI to strictly output the final result in English.
# Prompt to guide the AI to create perfect meeting minutes
system_prompt = """
You are an elite executive assistant and meeting minutes expert.
Analyze the provided transcript and create a perfect Markdown meeting minute.
CRITICAL: You MUST write the final output in English, regardless of the audio's original language.
[Meeting Minutes Format]
# ๐ Meeting / Lecture Summary Notes
## ๐ Core Topic (1-Line Summary)
(Summarize the core topic in one line)
## ๐ Core Discussions
(Detail 3-5 key points with bullet points based on the transcript)
## ๐ก Conclusions & Insights
(Final conclusions or key insights from the session)
## ๐ Action Items
(Action items clearly stating who, what, and by when. Or review points for lectures)
"""
print("๐ GPT-4o is structuring the transcript into professional Meeting Minutes...")
try:
# Generate the meeting minutes using GPT-4o
response = client.chat.completions.create(
model="gpt-4o",
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": f"Here is the transcript. Please create the meeting minutes:\n\n{full_transcript}"}
]
# Prompt to guide the AI to create perfect meeting minutes
system_prompt = """
You are an elite executive assistant and meeting minutes expert.
Analyze the provided transcript and create a perfect Markdown meeting minute.
CRITICAL: You MUST write the final output in English, regardless of the audio's original language.
[Meeting Minutes Format]
# ๐ Meeting / Lecture Summary Notes
## ๐ Core Topic (1-Line Summary)
(Summarize the core topic in one line)
## ๐ Core Discussions
(Detail 3-5 key points with bullet points based on the transcript)
## ๐ก Conclusions & Insights
(Final conclusions or key insights from the session)
## ๐ Action Items
(Action items clearly stating who, what, and by when. Or review points for lectures)
"""
print("๐ GPT-4o is structuring the transcript into professional Meeting Minutes...")
try:
# Generate the meeting minutes using GPT-4o
response = client.chat.completions.create(
model="gpt-4o",
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": f"Here is the transcript. Please create the meeting minutes:\n\n{full_transcript}"}
]
# Prompt to guide the AI to create perfect meeting minutes
system_prompt = """
You are an elite executive assistant and meeting minutes expert.
Analyze the provided transcript and create a perfect Markdown meeting minute.
CRITICAL: You MUST write the final output in English, regardless of the audio's original language.
[Meeting Minutes Format]
# ๐ Meeting / Lecture Summary Notes
## ๐ Core Topic (1-Line Summary)
(Summarize the core topic in one line)
## ๐ Core Discussions
(Detail 3-5 key points with bullet points based on the transcript)
## ๐ก Conclusions & Insights
(Final conclusions or key insights from the session)
## ๐ Action Items
(Action items clearly stating who, what, and by when. Or review points for lectures)
"""
print("๐ GPT-4o is structuring the transcript into professional Meeting Minutes...")
try:
# Generate the meeting minutes using GPT-4o
response = client.chat.completions.create(
model="gpt-4o",
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": f"Here is the transcript. Please create the meeting minutes:\n\n{full_transcript}"}
]
Step 5: Download to Your PC (.md File Save)
Did you see the chaotic conversation perfectly organized into 'Core Topics', 'Core Discussions', and 'Action Items' in the terminal?
Now it's time to download these flawless meeting minutes to your computer and share them with your team.
Conclusion
In the past, after a 1-hour Zoom meeting ended, the youngest employee acting as the scribe had to listen to the recording repeatedly and spend over 2 hours organizing the minutes. However, by using the 'AI Secretary Pipeline (Whisper + GPT-4o)' we built today, you can distribute perfectly formatted meeting minutes to your team in the time it takes to brew a cup of coffee (1~2 minutes).
If you are a college student, try uploading your professor's 2-hour major lecture recording. An incredible 'Summary Note' for your exam prep will be born.
AI is no longer just a toy; it is the absolute best weapon to help you leave work early. Bookmark your Colab link and use it in your meeting tomorrow! a2set Tutorials will be back with more of the most practical AI workflows.
