Speaker identification / diarization?

[sibbl]

Are there any plans to bring speaker identification into the project?

By letting users input multiple voice snippets, a speaker embedding can be created.
These embeddings can be used to recognize the speaker.

I'd like to replace my Google Home setup with HA Voice PE.
When I talk with it, it uses my calendar, my notes, my spotify account, ... and if my wife uses it, it uses her accounts.

Similarly I'd like to be able to assign different Assist instances to different HA users.
Everyone should be able to define their own available set of entities, actions, MCP servers, ...