I’m trying to get my head around why DQ skinning never took off? for example I just discovered Epic dropped support in UE4.
Anyone here got a DQ pipeline or is Linear still the flavour?
chur
Jamie
I can’t speak for others, but for our game on current PC hardware we considered the transformation cost too expensive to implement either on the CPU or in a shader. Also, DQ in general looks much better, but there are cases where it makes things look worse. For example, pinching and collapsing turns into hose-piping. Elbows should do neither. To solve this we would want to run both systems and let the artists author (texture map, vertex colors, etc.) how the two blend together. So, you’re multiplying your transform cost very roughly by a factor of 4. We would rather have more characters on screen than fewer better looking characters.
EDIT: Oh, and you can achieve the same quality with transform driven blend shapes. That is, author a good looking bent elbow, then create an additive blend shape that is driven by elbow angle. This kind of system also lets you do muscle bulging, etc.
Yeah true… certainly DQ suffers from deformation artifacts… I would suppose you could also place corrective blends on areas subject to bulging etc… I’m also wondering if the uptake of DQ by artists has been effected by a lack of education… i.e. just being taught the same old method year after year… I recently introduced DQ skinning in one of my classes but in effect I’ll have to ‘unlearn’ that class due to a lack of support in main stream engines…
Crytek have a paper here talking about their implementation of DQ… however I can’t confirm if that’s still the case.
On another note there is this paper on Elastic Implicit Skinning that appears to have some promise…
Certainly for now linear is the go to method. I’d like to see some kinda progress in this area in the future thou.
J
You could probably automatically pre-cook some number of elastic implicit skinning derived corrective blends at character creation or load time and then apply these in a linear weights system at runtime. It’s not as good as full runtime EIS, but it could solve most of the issues you might care about, and it has effectively 0 man-hour cost in production. Hmmm…
zero hours is always good
CryEngine is purely DQ skinning, no linear support in there.
I much prefer DQ skinning - it eliminates the need for several classes of helper bones and it’s pretty cheap to implement. It has artifacts – but linear does too. It’s just different artifacts.
Never underestimate the role of troglodyte I-don’t-want-to-learn-ism in these changes 
Do Max / Maya come with DQ skinning tools out of the box?
If not it simply represents an extra cost.
it’s built in to maya 2011 +
Both need extra helpers for proper deformation in the end, where DQ falls short linear is the go-to method and where Linear falls short DQ is the go to method. Talking with friends who have worked at Crytek and with CryEngine before they all say the same that in the end the work load for the tech artist was the same.
A proper solution would be one that allows for us to pain the blending between linear and DQ which would make for very slow calculations for games. Such blending exists for us already though in most 3rd party programs.
Yeah I’d like to see DQ skinning made available again…
for example…
http://www.jamietelford.com/binfour/DQ_skinning_Eye.webm
This eye uses DQ skinning for effecting eyelid movement… linear skinning won’t cut the cheese…
anyhow… DQ skinning is dead! long live DQ skinning!
We used DQ skinning successfully in a couple of games. The crotch and armpit areas were the hardest to get looking right, as they tended to balloon out with too much geometry.