Hey guys! so we are looking into this an have some way forwards. Thanks for all the feedback. We do want to provide you the most accurate summaries possible regardless of limitations so do allow us some time but you should see gradual improvements thanks!
You folks have been doing an amazing job keeping up with the deluge of suggestions, requests and complaints.
I have recorded and re-recorded my voice, but neither I nor other speakers are identified correctly. Often, some the speaker’s comments are attached to another speaker. Multiple speakers are labeled the same.
1 Like
This is a recurring problem for me as well.
hey guys! given current transcript generation limitations we are trying to solve this issue to the best of our ability so if you guys do not mind sharing the details of the recording where you faced this issue to hey@heypocket.com. If would give us more data points to work with so we can make this the best experience possible for everyone.
thanks, mihir
1 Like
thanks to @sidd_heypocket we were able to make some steps towards improving speaker identification in summaries! if you regenerate your summaries you should find the tagging a lot better thanks
2 Likes
First time user here (just got mine today!) After tagging speakers I regenerated the summary, and the summary still says “Speaker” and “Participant” rather than summarizing using names. I know diarization is very difficult, and the LLMs don’t do this out of the box. So I’m not surprised there’s occasional mixup. Just wanted to report my experience.
hey @RichTatum shouldn’t be happening @chay will take a look at this
1 Like
Hey @RichTatum
I am looking into this issue, and once it’s done will let you know here
Thanks for reporting!
1 Like
I also experienced this today. Particularly after tagging speakers and merging speakers and then regenerating. I had to go back to the first summary and then it said speaker 0 and speaker 3 instead of names. It was a mess