Geeks are Sexy Technology News

The main problem with photorealistic CGI humans, I’ve heard is ‘dead eyes’. That’s when the eyes just focus and lock on whatever the character is meant to be looking at.

Real eyes are naturally alive with ‘sparkles’ and move at an alarming rate. I don’t mean from consciously left to right and back again like watching a tennis match, but in terms of subconscious fractions of a millimetre every fraction of a second, constantly checking out the environment.

It’s almost impossible to recreate this in CGI. I think I’ve heard (exact statistics probably a little wrong, but close) that if Shrek’s eyes were ‘photoreal’ in the very first scene of the first film, that’d take 100x longer to render than the whole first film combined.

Not only this – even if the computing power WAS realistically possible for a company to fork out for, it’s near impossible to work out where the eyes would focus on in an environment, since the whole thing works on such a subconscious level at such minimal and fast rates.

So however great the textures, face shapes and muscles might look, the eyes are always what gives the game away.

The main problem with photorealistic CGI humans, I’ve heard is ‘dead eyes’. That’s when the eyes just focus and lock on whatever the character is meant to be looking at.

Real eyes are naturally alive with ‘sparkles’ and move at an alarming rate. I don’t mean from consciously left to right and back again like watching a tennis match, but in terms of subconscious fractions of a millimetre every fraction of a second, constantly checking out the environment.

It’s almost impossible to recreate this in CGI. I think I’ve heard (exact statistics probably a little wrong, but close) that if Shrek’s eyes were ‘photoreal’ in the very first scene of the first film, that’d take 100x longer to render than the whole first film combined.

Not only this – even if the computing power WAS realistically possible for a company to fork out for, it’s near impossible to work out where the eyes would focus on in an environment, since the whole thing works on such a subconscious level at such minimal and fast rates.

So however great the textures, face shapes and muscles might look, the eyes are always what gives the game away.

I disagree, it’s not the just the “sparkle”. I may be off base to say this but it seems to me that all the effort to capture the surface information of the actress’ face was for naught if they just apply it to a less than perfect puppet structure. I think the incongruity that we are perceiving is from the unrealistic movement of muscles and form behind the face. If they could take a detailed map of the real muscles under the actress’ skin and then animate within the parameters of those real muscles, it probably would achieve a closer, more realistic facsimile. How would they get the detailed muscle structure without pulling her face off or How much computational power would it require to simulate all the muscles of the face at once are questions i’d like to know answers to.

I disagree, it’s not the just the “sparkle”. I may be off base to say this but it seems to me that all the effort to capture the surface information of the actress’ face was for naught if they just apply it to a less than perfect puppet structure. I think the incongruity that we are perceiving is from the unrealistic movement of muscles and form behind the face. If they could take a detailed map of the real muscles under the actress’ skin and then animate within the parameters of those real muscles, it probably would achieve a closer, more realistic facsimile. How would they get the detailed muscle structure without pulling her face off or How much computational power would it require to simulate all the muscles of the face at once are questions i’d like to know answers to.