Never take meeting notes again
Never take meeting notes again
Or why you might want to invite big brother on to your laptop
I don’t know how many meetings I’ve been in where people are busy taking notes but not actually (and actively) listening. What if you could have faith that every meeting was accurately captured, summarized, and action items identified? What if you could even have the tone of the conversations analyzed to take a temperature of how the meeting went?
While some products like Microsoft Teams have real-time meeting recaps as a Premium option, it often comes with constraints, additional payments, or doesn’t work if the meeting is with an organization that doesn’t use Teams. I’ve created a pipeline that works agnostic of the video conference technology platform or audio source for the cost of a few cents.
While the steps might seem a little daunting, it’s really quite simple even for a non-techie. Once set up this literally takes a minute each time you need it in the future.
Audio Capture
First of all, you need to capture the audio. You should know that the laws regarding audio capture vary by state and by country. This is not legal advice, but If you’re in the US, you can check here for whether you’re in a one-party or all-party consent state. Regardless of the legal obligation, it’s just good form to ask the participants and respect their choice.
Assuming all parties have consented (be sure to send them the notes too!), then a great open-source tool for capturing any audio (and video) on a PC, Mac & Linux is OBS Studio.
Download OBS Studio and ensure your earbuds/headset is connected.
Launch OBS Studio and Click on “Settings”.
Select “Audio” from the left side menu and click on the dropdown next to “Desktop Audio”. Pick the headphones/earbuds you use to listen to a call as well as for speaking by selecting “Mic/Auxilary Audio”.
Click on “Output” on the left-hand menu and set the ”Recording Path” to wherever you want to place the captured conversation audio file, such as the desktop.
Click ”Apply” and ”OK” when done and you’ll now be at the main OBS Scene.
The general flow works from Left to Right here. First, we’ll create a ”Scene” which is really like creating a small project that contains all our settings. Then we’ll select which source (The Video Meeting Window) we want to capture from. The Audio Mixer will show levels once you’re in a conversation to know that audio is being heard and captured, and to the far right you can Start / Stop the recording itself. I’ll walk you through this below.
Clicks the ”+” icon in the bottom left of the screen and enter a Name for this scene, e.g. ”Video Meeting Recorder”, and click ”OK”. You’ll only need to do this once as it’ll remember your settings next time you come in. You should now see it listed and highlighted under Scenes.
Click on the ”+” icon in the lower left of the ”Sources” section and select ”Application Audio Capture (BETA)” from the top of the list. It’ll prompt you to create a new Source and default to ”Application Audio Capture (BETA)” which you can just select with ‘Ok’
Now you’ll select the meeting window that you wish to record. Typically this’ll be obvious by the browser tab name and can be selected before you ”Join” the meeting itself. You may need to play with this a bit to ensure you’re recording the right window, but conceptually you’re just specifying the source Chrome window or application to be recorded.
In my case, I usually use Google Meet, and the window name for my example meeting I have open is ”Meet - zhb-vyfs-vnm - Google Chrome”. Select the window and hit ”OK”.
Assuming you’re in the lobby of the meeting. You should now see the levels for your Mic start to pickup any noise. Once the other participants are speaking you should see their audio appear in the ”Audio Mixer” levels area.
To start and stop the recording simply select ”Start Recording” before you enter the meeting or during it, and ”Stop Recording” when you’re done. The resulting audio file will be stored wherever you specified earlier, usually with a timestamp filename like, ”2023-10-14 12-29-09.mp4”.
Transcription
Congratulations! You now know how to capture audio from your desktop without being dependent on any paid software or your audio being stored in the cloud.
Now let’s transcribe the audio file into something that we can feed to ChatGPT in a textual form. I like another great free utility called ”Buzz” by Chidi Williams.
Download ”Buzz” from https://github.com/chidiwilliams/buzz
Select the latest release and download the ”Asset” for your Windows or Mac machine.
2. Install and launch the ”Buzz” software and you’ll be greeted with an application window that looks like this.
Select ”Help” then ”Preferences” and you’ll see it asks for your OpenAI API Key. Enter your ChatGPT API key and click ‘Test’ to make sure it’s working correctly.
Note: It’s very simple to get your API key, simply follow the instructions at.
https://help.openai.com/en/articles/4936850-where-do-i-find-my-secret-api-key
As of writing, you can get a free trial of the API Key with about $18 of credit, but if you want to continue using it then you’ll have to pay a small nominal fee. More details here. https://openai.com/pricing. For our purposes, we’re usually talking about a few cents per conversation.
Click the Models tab and select ”Tiny or Small” and select ”OK”. This is essentially downloading the OpenAI Whisper Speech to Text model to your computer where it will process the audio file locally to create a textual transcript file. I’ve tried the ”Large Model” but found that the transcription time was much, much longer, and not much better in terms of quality, but your results may vary.
We’re now ready to import the audio file of our video meeting and create a transcription! Go to ”File”, select ”Import Media File”, and select the audio file we created using OBS Studio.
You’ll now confirm the settings as shown below and hit ”Run”. The resulting text transcription should be placed in the same location as your source audio file.
Note: For an hour-long conversation this might take 15-20 mins to transcribe and will be dependent upon your machine’s hardware. You’ll see a progress % update along the way. The software doesn’t currently do speaker diarisation, which is a fancy way of saying that it doesn’t distinguish between different speakers in the conversation but ChatGPT seems to figure this out automatically. As you’ll see below, I’ve not found this to usually not be a problem, especially in 1:1 conversations.
Summarize, Analyze and Share
OK! We now have an audio capture of the meeting and a usually very accurate textual transcription in a text file. Time for ChatGPT to do its thing and process the text into something useful!
Open up a ChatGPT session at
https://chat.openai.com/ and select GPT-3.5 or GPT-4. I’ve found good results with both.
Give ChatGPT context with a prompt like:
Imagine you’re a professional transcriber, I’ll provide you with a transcript of a meeting between myself and Claire from accounting. I’d like you to create meeting notes in the following format. A highly detailed, verbose summary, call out any and all action items that were discussed, brainstorm 5 additional ways we could help one another and point out any discussion points that might have been confusing or unclear. In addition create a ”tldr” summary at the top, along with details of the time, location and sentiment of the speakers
ChatGPT will respond with something along the lines of, “Absolutely, Adam. Please go ahead and provide the transcript of the meeting with Claire from accounting, and I'll create the meeting notes as requested.
“Open your audio transcription text file, and hit ”CTRL/CMD-A” to select all text, ”CTRL/CMD-C” to copy the text.
Paste the full transcript into the ChatGPT Window. The results will be a concise and detailed report that succinctly summarizes the conversation with specific actions and recommendations called out.
Here’s an example based on a fictional conversation I asked ChatGPT to create, in terms of what you might expect:
TL;DR Summary
Time: October 14, 2023, 2:00 pm - 3:00 pm
Location: Conference Room A, Corporate Office
Sentiment: Collaborative but cautious; both Adam and Claire are keen to align, yet aware of the challenges ahead.Detailed Meeting Notes
Agenda
Introduction and setting the stage
Review of the current year's budget
Planning for next year's budget
Pain points and challenges
Discussion of growth opportunities
Final thoughts and action items
Introduction and Setting the Stage
Adam and Claire acknowledged the importance of budget planning for the upcoming fiscal year. Adam emphasized the need for agility and innovation, given his entrepreneurial background and tech focus. Claire, representing accounting, expressed concerns about maintaining financial health and compliance.
Review of Current Year's Budget
Adam pointed out that the technology budget was almost exhausted due to unexpected needs in cybersecurity.
Claire highlighted that although some departments underperformed, Accounting was able to reallocate funds efficiently.
Planning for Next Year’s Budget
Adam suggested a 10% increase in the tech budget to accommodate emerging technologies.
Claire advised a more conservative approach, recommending a 5% increase with a contingency reserve.
Action Item 1: Adam to draft a detailed proposal for the suggested 10% budget increase.
Action Item 2: Claire to provide data on this year’s spending to help inform next year’s budget.Pain Points and Challenges
Adam mentioned the difficulty in predicting costs for adopting new technologies.
Claire talked about challenges in financial reporting due to remote work.
Action Item 3: Both to meet again in two weeks to review progress and adjust plans accordingly.
Discussion of Growth Opportunities
Adam emphasized the potential ROI from investing in AI and Cloud Computing.
Claire was interested but asked for data to back up these claims.
Action Item 4: Adam to compile ROI metrics for proposed tech investments.
Final Thoughts and Action Items
Both agreed that a more detailed follow-up meeting is needed.
Claire to lead the financial review and Adam to spearhead tech strategy.
Action Item 5: Schedule a follow-up meeting for a deep dive into specifics.
Brainstorm: Five Additional Ways to Help One Another
Cost-Benefit Analysis Workshop: Adam could host a session to help the Accounting team understand the value of tech investments.
Financial Literacy for Tech Teams: Claire could conduct a seminar to educate tech teams on budgeting and financial planning.
Monthly Sync-Ups: Regular meetings to ensure alignment and adapt to any changes quickly.
Tech Audit: Adam could help Claire understand what software/tools could make Accounting more efficient.
Resource Pooling: Explore opportunities to share resources between departments for mutual benefit.
Points of Confusion or Unclear Discussion
The exact percentage of budget increase remained unresolved, requiring more data and discussions.
The implementation roadmap for new technologies was not clearly defined, leading to some ambiguity about timeframes.
So there you have it. I appreciate this might seem a little daunting at first, and undoubtedly there are paid tools and services that can replicate much of this. But I like it because:
It’s essentially free or costs very little
It allows me to easily capture any audio or conversation and feed it to ChatGPT to discuss further.
It only takes me a few seconds to start the recording after gaining consent and capture the audio. Another few seconds to drag and drop it to ”Buzz” afterward and another few seconds after transcribing to copy it into Chatgpt.
Conclusion
As an information worker, I consider this a MUST-HAVE capability whether you cobble it together yourself or use a paid service like Otter.ai . In a future post, I’ll detail more about how I integrate this method into my Personal Knowledge Management workflow using ”Obsidian”.
Hopefully, this will become second nature, and your colleagues will thank you for taking some of the best meeting notes they’ve ever seen.
Stay ahead and let Sheppard Consulting help navigate your business through the new technology era. Let’s Talk.