Fediverse

I also share the privacy concerns - it'd be too high a bar of publicly identifying yourself to really encourage conversation (and then I'd be worried about the perhaps non-technical people who would accidentally use it not realizing how their voice is PII).

I'm intrigued by the idea from a UX/convenience perspective though. Maybe you can get the best of both worlds (pseudo anonymity + speak-to-participate UX) if you "salt and hash" the audio consistently? That is, every user ID would have their voice altered in a consistent but irreversable way, that's also not vulnerable to some future rainbow table~esque attack.