It's partly a question of dataset, but also a question of how similar the voice "code" is in the latent space. Maybe they picked that point to be similar intentionally, or it was an accident… either way, if it's similar moral rights and Right Of Publicity are infringed.

→ View original post on X — @alexjc