Would like to differentiate different voices in the transcript.
Wil
I think the development issue is Hedy is not actually "listening" to voice, tone, volume or anything else other than pure words. I see the benefit of adding this, but worry about if API costs for Hedy outweighs the actual benefit. A slight modification in my own dialogue, as I highlighted above, has drastically improved this. Hedy no longer says "The speaker said" and uses proper names.
Wil
My current workaround to this is to use names in my conversations more and if someone says a highlight, I say something like "Great Feedback Tim" and Hedy does pretty good at keeping context aligned with that person. I've tested this by asking person specific questions and it was correct every time.
Canny AI
Merged in a post:
Speaker Identification
C
Carina Clingman
I use descript for video editing, and it can detect different voices and segment the transcript to identify speakers, then I add labels to name each speaker. That is very useful for creating accurate meeting notes where I can attribute ideas and questions to each person in the meeting.
Canny AI
Merged in a post:
Multiple speaker detection
S
Sebastien R
Not sure how some meeting recorder manage to detect multiple speakers when not being added to a meeting, but it would be an awesome feature
A
AJ
Can we merge this idea with the one provided below which already has nearly 50 votes so that we can merge vote count?
A
AJ
Speaker diarization is an important update in the recording process for Hedy, as it enables accurate attribution of action items and other key points to specific individuals involved in the meeting. This helps to ensure that follow-up discussions or decisions can be effectively communicated to the relevant parties.
Julian
Merged in a post:
Speaker recognition identification
N
Nghe
Please add a speaker recognition option for the transcription. Perhaps using certain thresholds, you can identify speakers as Speaker 1, Speaker 2, etc. At a later release, it would be even better if we can add custom labels to each speaker if we happen to know who said what sentence.
S
Sebastien R
Or during the live transcript using different colours. We could then associate a name to each identified persona.
With a click we could correct which colour/persona mentioned an output, in case of errors.
Also, give live statistics about who spoke the most etc.