group call listening idea

[humble igloo]

Whenever neuro is in a group call, neuro will barely talk unless spoken to, and even when spoken to she wont talk because everyone else is still talking.

So my idea is to have a mode for neuro to listen to each person seperately, but only talk if her name is called or maybe mentioned.

This would still need the person who mentioned neuro to stop talking but is probably better than the whole group making no noise.

Thoughts?

[sour violet]

my idea is to have a mode for neuro to listen to each person seperately
Wouldn't this require for all those people to install a custom program that only listens to their mic and transfers it to some processing system? AFAIK distinguishing overlapping voices is not a trivial task

[patent mason]

Right now neuro doesn't distinguish between speakers when she's in a group call, which is why she kept misnaming people in the filian dating show and called one of the contestants in Parasocial mizkif. I like the idea, but neuro needs to be able to distinguish between voices first, and there's multiple ways vedal could go about doing this. It's probably something we will see being implemented soon enough

[amber glen]

Theres voice recognition software that recognizes the tone and pitch of a voice rather than just the words spoken, and with that can tell "users" apart. That could be the solution to it, but i imagine thats a much more costly solution.

[humble igloo]

yes, but I imagine something like everyone is on 2 discord calls one for the whole group and one just with neuro

or something

[sour violet]

I can see it working on a technical level, but ah.. people are so lazy :D

[humble igloo]

understandable

[rapid linden]

there's actually an easier option to any of that tbh, but i shall

[humble igloo]

no way

[dapper yarrow]

[maiden light]

Leaks? No way

[stuck halo]

you mean seeing who is talking from discord voice activity

[humble igloo]

I don't think she'll be able to tell which sound came from which if multiple people is speaking

[stuck halo]

in that case she won't be able to tell what words are being said anyway

[humble igloo]

yes that is why I suggested that everyone is on 2 calls, one with neuro only and the other one with the group

[thorn bramble]

I thought of Neuro being able to track who's talking based on dialogue and vision (when the stream starts, she will associate a certain voice to someone so that she doesn't get confused)

[humble igloo]

I think that would be the case in the future since that is quite advanced I think, so my suggestion would be like a temporary fix

[stuck halo]

thats a lot of work and would require multiple discord accounts - I would just use an app they you can extract the audio streams from separately

also consider that running voice recognition costs $

[humble igloo]

Yes lets grow this idea more by discussing it, and the money problem.. Well Vedal is a Billionaire

: True

[stuck halo]

eccentric one yes but i don't know if that money is in cash, its probably stuck in assets

I found this, but they look paid or recording only
https://www.reddit.com/r/obs/comments/gposa5/how_to_separate_discord_user_voices_on_different/
I'm gonna look at the discord internals and check something

it looks like skype would be better for this tbh

actually i wonder

[humble igloo]

[stuck halo]

it might be possible to unpack the webrtc aecdump live - idk tho

[humble igloo]

I have no idea what you just said

now we wait

[amber glen]

You want to turn out little Neuro into a boomer who uses Skype in 2023?

[stuck halo]

yeah and she can send 300 MB files without nitro