There's a lot of subtle things that can get things caught in the uncanny valley that you really wouldn't think of naturally - though I suppose that is why the uncanny valley is triggered. For example, in most cases, you'll be able to intuitively tell if a face is CG'd the moment it starts speaking - why? Pay attention to the small things, sometimes it'll be something like the upper lip movement - try moving your upper lip it's impossible to move it in isolation. Furthermore, pay attention to lighting, well not the lighting in particular, but how the light interacts with the face model - is there subsurface scattering - do certain parts of the face redden/self-illuminate in response to light, does the skin "sparkle" over the pores, do small hairs cast shadows. You can go deeper with how you map the face as well with various bump maps to break up the surface so that it is not completely smooth, how the pores of the face stretch. One easy "cheat" to give an extra bit of "oomph" to some CG characters is to place extra attention on how the eyes are lit and place an highlight in the eye that doesn't/can't realistically exist to give more "fake realism."

That all being said, could animating photoreal CG faces die off as deepfake tech gets better and more optimized? Disney's already looking into it.

^^ wow that's fascinating! I'm curious how we might implement a shader node, or combine multiple shaders to model the small facial hairs, subsurface scattering/skin translucence, and pores most accurately ~ and finding a balance between realism and cartoon-like style to tell a story

Thanos is a pretty cool example of a character with a CG face that used motion capture and a combination of a lot of other methods. They modeled his face down to the pores and even added some hairs to his face which helped make it look more real.

Really cool <1 minute example of this: https://www.youtube.com/watch?v=HjHiC0mt4Ts . Note how the subtle things really do add up to create a feeling of both ultra realistic and unrealistic.