Recording a long phone call, did a great job of identifying the two speakers, transcribing, and summary. Recorded a presentation to a Board meeting, again very good summary.
one thing i was hoping for is for me to reassign a speaker and than all the following or subsequent people would be changed - but that doesnât seem to happen. As an example if i change [Person 1] to Jason - the next point still lists [Person 1] and thereâs no way i can go through each conversation and correctly tag everyone. Once I change [Person 1] to Jason - the system should automatically change all [Person 1] in that conversation to Jason.
Hey @kp1 when you rename a Speaker X all instances of those speakers are already renamed if thatâs something you are not seeing please gimme a message! Iâll check it out for you. Ideally whatâs happening is the ai must have labeled the segment as a different speaker
My primary use is for Work meetings (zoom calls), and I am comparing meeting summaries from Notion, Granola and Pocket and so far Pocket is coming in at #3 in that list. I am still going to give it some time cause i think Pocket has the advantage of also using it for whiteboard sessions and in-person meetings (and I am rooting for the device!), but few things need to really improve:
- The transcription itself is buggy, it is not going a great job in identifying different speakers and when transitions happen. I have limited time to manually correct the transcription each time.
- The summaries have 10s of options, but i just want âAuto-detectâ or âAuto-pilotâ to work, pick the most suitable theme and give me a good summary (vs everytime I have regenerate and it seems like an experiment).
- The summaries are not yet at quality where I can share with others (Granola and Notion do a better job at generating these summaries so far).
Iâve been using Notion for meetings for several months. Iâve found it has speaker identification issues for me. While it does recognize that there are different speakers, it has a weird habit in one-on-one meetings (my most common use) of reversing them. Honestly, I think speaker identification is a weak area in every AI transcription system Iâve tried. I, too, am rooting for Pocket to continue to improve in this area.
Thereâs a least one thread floating around on the best model/summary combinations. If you havenât read it, you might want to.