Hi there - just got my Pocket a few days ago. First few days of using it has been great. I am trying to edit Speakers in the summaries and for the life of me I cannot figure out the UX to label speakers. I see “Speaker 0” and “Speaker 1”, but there’s no workflow for editing them and having it flow through the rest of the summary chunks.
I’ve read a few confusing posts here about the state of speaker identification and tagging. Is this a known bug, or by design at the moment? Thanks in advance for the help.
Speaker labeling/tagging is almost certainly the buggiest part of the app still.
This is a pretty huge problem at the moment. I just ran into yet another issue with the current workflow:
Issue: Merging misidentified speakers and renaming them incorrectly updates ALL past memories with the new name - even conversations where the speaker never appeared. A 2-person conversation was detected as 4 individuals; after merging and renaming, unrelated past memories were corrupted with the wrong speaker name.
Root cause: Speaker merges link to global speaker profiles instead of staying scoped to the current conversation. Renames propagate across the entire memory database.
Potential solution: Scope merges to the conversation level and require explicit confirmation before linking to existing speaker profiles.
@dps but speakers profiles are scoped globally so that we can map users across conversation. This isn’t something that is an issue when, we identify a user we automatically attach the same.
We can figure out a way to change the merge behaviour but something that will be prioritized next week mostly.
If there are any issues you see please let us know but when it comes to the core functionality we haven’t gotten any more complaints always open to looking at something if there is anything wrong
hey @michael2000 to rename speakers you can go to view transcript in your summary screen and edit names from there. its in the 3 dot menu on the top right!
just something I want to add here if there is any other functionality change that you would like in the speaker renaming/tagging, transcript edit we would love to hear it here so we can prioritize accordingly
Understood on the global profile design. That makes sense for mapping users across conversations. The issue is we currently have no way to say “this speaker was misidentified; link this conversation to a different existing profile instead.”
Right now, tapping “Rename Speaker” looks like reassignment but actually renames the profile globally, which corrupts past memories when the AI misidentifies someone. We need a separate action like “Link to Different Speaker” to reassign conversations without changing the global profile name.
I would love to understand how you would live this to be shown. Do you also want us to show which conversations this speaker is linked too? So these decisions can be made accordingly?
“Adding Speakers” should add or remove global speaker profiles from this memory (not creating a new profile). “Rename Speaker” should reassign the current speaker to a different existing global profile, updating all segments in this memory. Merging separately identified speakers should require you to select which global profile to merge into, resulting in a single identified speaker linked to that profile.
@mhirrrr isn’t part of the challenge that you have to name the speaker the same way every single time?
I’ve been naming them differently depending on who the audience is for the notes and I’ve realized watching feedback on this and the Pocket’s team input - I may be creating my own problem ha.
It must be done on the app. In the top of the transcript view there is a person icon which allows you to view all speakers. On the right of each speaker there are 3 dots, click this and you can then ‘edit speaker name‘.
When viewing the transcript, you can also tap on any already labeled speaker to update it to be a pre-existing speaker.
I understand the reason for the global speaker concept, but without the ability to say, “you were wrong, it’s this other person,” I’m getting a TON of mistagged transcripts.
I’m now also getting an issue where multiple speakers are getting tagged as me. My name shoes up multiples times in the speakers list. If I tried to edit one of them to the actual speaker, it actually changes both. See video below.
hey @chucksense makes sense we have taken @dps feedback into account. We are working on making those changes as soon as this weeks release cycle is closed