Recognizing someone has everything to do with hours of exposure. Which is why people are better at picking their mother out of a crowd rather than a random person who they were shown a picture of once.
I'm beginning to wonder if you've ever considered your interactions with other people, or if you're just making this up as you go along.
We certainly have memory, and can forget people, like a random person we were once shown a picture of. And pictures are not 3D faces either.
But if you meet someone for half an hour and then run into them later that day, you recognize them just as easily as you do your mother. There's zero doubt that they are the person you met earlier (unless you have face blindness or something). You don't have to meet them for 10 more hours to be able to recognize them instantly.
Sure you might forget them a year later but that has nothing to do with your recognition capabilities, but with memory.
Ok, and how often does an average person hear Scarlett's voice that they would recognize her voice when they hear a random voice? This isn't a situation where someone hears a Scarlett voice sample, and then compares it to Sky. This is, someone has heard Scarlett randomly in movies over the past 20 years and then recognizes Scarlett as Sky.
Which, given the fact that the voice actor for Sky has put out a statement stating that her voice has never been compared to Scarlett's in real life, makes me think that people are just hallucinating that Sky sounds like Scarlett. For what it is worth, I listened to them side by side(before OpenAI put out a statement by the voice actor) and they are not the same voice.
I'm beginning to wonder if you've ever considered your interactions with other people, or if you're just making this up as you go along.