"You won't be able to tell," said Hao Li, a leading researcher on computer-generated video at USC who founded Pinscreen in 2015. "With further deep-learning advancements, especially on mobile devices, we'll be able to produce completely photoreal avatars in real time."