The most extreme real-world usability testing took place in full board meetings of fourteen people, and occasionally with additional members of the public. It’s about as chaotic a situation as I could imagine for the AI-powered recorder, with advisors frequently interjecting and talking over each other, some of whom lack public speaking skills and have barely audible voices. We’re all seated at different tables, and the farthest distance from one speaker to another is about 13 feet.
I’m happy to say that with the Plaud Note Pro placed in a central location during the meeting, it performed like a champ. My old workflow was to record on my iPhone (longest meetings typically last around two hours), upload it to Fireflies.ai, then feed the transcript into ChatGPT Plus to summarize it with a little nudge to correct obvious errors or misidentified speakers.
It’s clear that the four microphones are doing a lot of work here to enable more accurate transcriptions. I compared it with some of my previous recordings, and the difference was night and day; I could easily identify the speakers rather than guessing someone mumbling something from across the room. The words were clear, so the AI guessed less. That’s the real selling point here.
Once you label speakers once, Plaud remembers their voice patterns for the next time, so each recording becomes more accurate and requires less effort on the part of the user. The custom vocabulary feature makes it easier to recognize local place names. Therefore, if you find something wrong regularly, you can also enable it.
There is another powerful aspect to the software side of Plaud that would be easy to overlook. Once a recording is completed, you don’t need to ask them to transcribe it, then summarize it, and then email the relevant parties a list of actions to take. You can create an automation to do everything at once, and with the latest firmware, you don’t even need to open the app anymore: just bring your phone close enough to detect the device.
Automations require a keyword to be issued within a customizable time frame; as a “staff meeting” in the first five minutes. Once identified, the trigger will execute and your reports will be generated. In practice, our meetings are filled with lots of inconsequential nonsense that shouldn’t be recorded. So I still wanted a monitoring step before a transcription was automatically emailed.
