The Next Gen of LA Noire Facial Animation Technology

[xTheSweetyMan]

your *doom*ed!

i smell a cameo!! :D

[HappyGuy00000]

your doomed !

[trugangsta4real]

WOW!

[tokyopp]

awsome video but what is in the background?

[12WaterPhoenix12]

@SECx8xProductions It wasn't rockstars first try. Team Bondi made the tech and the game. Rockstar was just hte publisher.

[improvisedcinema1]

looks like a stereotypical gears of war character

[CadenBurleson]

BAD-ASS

[mr2000jp]

shocking brilliant

[tohtorigyro]

didn't they use both techniques combined when making avatar?

[DanFrederiksen]

it's a mistake to place fiducials. just have the camera see the face, fiducials only degrade the data, not enhance it.

[gopofk]

at first i told this is fake :D

[drofferni]

Muito bom!!!!!!

tecnologia surppreendendo cada vez mais

[R3Dpill33]

one prob with la noire is sleight lack of eye contact... i cant really explain it here cuz thats not the full prob

but its still good

[78097]

Janimation produce visual effects for film. It's impressive in a different way, and there are many reasons why the two should not be compared. The most notable of which is technical limitations: the above video is not real time. This does not mean that their work is without merit, I just cannot stress the differences in production enough. Team Bondi has produced some really ground breaking facial animation with incredibly tight technical restrictions.

[Vidgees]

@78097 Keep in mind, film CG technology and video game mocap technology are quickly merging. It's evident from LA Noire that locking in an actors head with 39 cameras filming the actor isn't the complete answer to the problem. LA Noire has some strange oddities that occur when people are talking to each other, but this to be expected when the actor isn't able to actually act in the scene, with the other actor he/she is talking to. The two technologies need to merge for the best results.

[stinky472]

@Vidgees I think the next-gen technology is a combination of the two as you said. 3d scanning methods obviously capture subtle nuances better: it doesn't matter how many tracking points you put on an actor's face or how well you simulate the muscles. It'd be like trying to mo-cap the tendons on the surface of a hand: it just doesn't get that granular. Then again 3d scanning cannot control an arbitrary mesh of, say, a fantasy creature, let alone a clean, high-res mesh of the same actor.