The team notes their approach is conceptually quite simple, but is able to reliably clone voices using dramatically fewer compute resources.