I run pipewire and let me offer some generic tips:
* Make loopback devices that are always there for headphones/mic -- software doesn't handle connection/disconnection well
* Keep all your processing in carla (i load it with systemd as a user unit on login)
* You still need some way to control "pulse audio" (even though it's pipewire provided) for per-app volume control, profile selections, etc, so you probably want pavucontrol for that.
This is interesting as there's a commercial product also called Lyrebird (https://www.descript.com/lyrebird) which does voice replication. Also I would expect a voice changer app of all things to have an audio demo linked prominently in the README so I can hear how it sounds!
From a quick read of the code, it seems like all this does is shift pitch; it doesn't change resonance or anything else that affects how your voice sounds. So it's great for sounding like a chipmunk or a demon, but not that useful for sounding like a different person.
Looks like it uses libsox internally. You can also use command-line SoX utilities to get the same effect. For example, this plays your microphone back at you, pitch shifted and with a delay:
(To be clear, I don't mean to beat down anybody's pet project. Just spreading related information. There's very little accessible documentation of Linux audio stuff.)