Nästa Generations Ansiktsdeformation: Ansiktsdeformation, leder, blendshapes med ett ekonomiskt och snyggt tillvägagångssätt
Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesis
When making a digital face look real, the hardest part is making it move naturally. There are hardly any isolated movements in the face, when you squint your eyes, the skin on your neck stretches. When you open your mouth your temples wobble in and out. Every movement is a reaction to something else happening. In movies you have complex muscle-simulations that pushes and pulls the skin, algorithms that slides the skin over bone markers, allowing polygons to move freely while remembering the shape of the silhouette. In movies this is not an issue because the simulations are fixed and perfected beforehand and can be processed over days (sometimes weeks) in massive renderfarms. In a game you have approximately 0.016 seconds of computation time for each frame, to hit the magic number of 60 frames per second. That is 0.016 seconds for all models, all animation and all textures etc, to be read and rendered.This puts us at a very strict budget of what we can and are allowed to do. So far the cheapest way of doing animation has been joint driven animations but the limitations are now becoming more and more apparent. The era of joints have passed and for realistic animation, joints can only give us the base from which to work and explore, but it cannot bring us to perfection in the same way that other more expensive methods allow.In this paper I will be demonstrating how the previous generation worked with joint driven animation, I will use an existing system at DICE, only driven by joints, and expand that to include blendshapes which will allows us to simulate skin sliding and more complex movement of the face that the previous method doesn’t allow. And how to make this is cheap as possible in terms of workload for the people using the system. I will also discuss the issues that comes with blendshapes and why a combination of both joints and blendshapes is still the best and cheapest way to go for the game industry.
Place, publisher, year, edition, pages
2014. , 17 p.
IdentifiersURN: urn:nbn:se:ltu:diva-45090Local ID: 2d2b97e5-1032-4739-aa61-75b3aa003d40OAI: oai:DiVA.org:ltu-45090DiVA: diva2:1018370
Subject / course
Student thesis, at least 15 credits
Media Design, bachelor's level
Validerat; 20140624 (global_studentproject_submitter)2016-10-042016-10-04Bibliographically approved