FA+
Shop
14540
Views
Views
391
Favorites
Favorites
Category
All / All
Species Cheetah
Size 640 x 360
File Size 447.9 kB
Report this content
★
More from zorryn
Early FaceRig Test of Verall
https://vimeo.com/108024374
The guys from FaceRig.com were so kind to turn on of my characters in a live-tracking-avatar :)
Check out this tool at facerig.com
The guys from FaceRig.com were so kind to turn on of my characters in a live-tracking-avatar :)
Check out this tool at facerig.com
Category All / All
Species Cheetah
Size 640 x 360px
File Size 447.9 kB
It's realtime fur, and as far as I gathered it is their custom engine. But from the look and appearance it is the old and known "shell"-effect, where a script creats multimple shells of the mesh and places dots per alpha map on each. Combined they look like volumetric fur.
Well, with bones you are limited to general directions like skaling, translation and rotation. With blendshapes, virtually every vertex has its own "bone". You are actually free to sculpt the expression with the vertices you have.
The only downside of blendshapes are their linear transition. For eyelids and jaws etc. you want a rotational origin. But for all parts of muscles putting strain on the face, linear deformation is good enough.
The only downside of blendshapes are their linear transition. For eyelids and jaws etc. you want a rotational origin. But for all parts of muscles putting strain on the face, linear deformation is good enough.
There's advantages and disadvantages to both. For the uninitiated, bones more closely resemble the major underlying structure of the body or things that don't flex, so a jaw bone, "eye bones", and to an extent the tongue and ears all make more sense to have rigged bones. They just move or rotate in specific ranges (like your jaw is a hinge, your eyes rotate mostly in the same place in your head, etc.). But, muscle deforms the skin over the unflexing parts, something bones don't do so well.
Add enough bones and rig everything right, and you can approximate muscle movement. However it's easier to make a smiling face for example with shape deformations that looks exactly how you want. Basically you start with a neutral face, and then take a picture of it in that state. Then change the positions of the vertices of the face a bit, pulling some this way or that until it looks like the character's smiling, or frowning, or angry. You take another picture of it in that state, and then tell the 3D software to blend between the neutral face and that other face. It allows for subtle things like partial smiles, and it's typically easy to set up.
I fully understand why they don't do it though. Shape deformation works in two ways. Either stuck between neutral and one of a set of other poses as above; like if you don't have a face for say, being scared, the face could only partial angry and/or partial happy, hoping to approximate what you're doing. OR the face's vertices can be morphed dynamically by trying to map the topography person's face to the topography of the mesh, in theory resulting in extremely accurate representations of your expression. The trouble is, the target mesh might be too low quality to map to appropriately and might have a strangely different topography as well. How do you map a muzzle to a human face for instance? Or cat ears on top of the head?
Current real time facial recognition techniques find common points on a face and use those as points of reference. For instance it's very easy to find the corners of a person's mouth, corners and centers of their eyes, their chin, end of their nose, and the outsides of their face. This is because (barring mutation or mutilation) we all have the same general appearance. If you can find things like a person's mouth and chin, you can determine how a jaw bone should move. You can usually figure out where the cheeks are as well, and using the outside corners of the eyes and the corners of the mouth, figure out how the facial muscles are moving on a general level (bunching up for a smile, drooping for a scowl, squashing and stretching for a :\ face, etc.). These motions can all be translated to bones which are pre-rigged to move certain vertices very easily on the face. At least, very easily compared to figuring out how to move all the vertices of a face, compared to an entirely alien face (like a human to a dragon).
Add enough bones and rig everything right, and you can approximate muscle movement. However it's easier to make a smiling face for example with shape deformations that looks exactly how you want. Basically you start with a neutral face, and then take a picture of it in that state. Then change the positions of the vertices of the face a bit, pulling some this way or that until it looks like the character's smiling, or frowning, or angry. You take another picture of it in that state, and then tell the 3D software to blend between the neutral face and that other face. It allows for subtle things like partial smiles, and it's typically easy to set up.
I fully understand why they don't do it though. Shape deformation works in two ways. Either stuck between neutral and one of a set of other poses as above; like if you don't have a face for say, being scared, the face could only partial angry and/or partial happy, hoping to approximate what you're doing. OR the face's vertices can be morphed dynamically by trying to map the topography person's face to the topography of the mesh, in theory resulting in extremely accurate representations of your expression. The trouble is, the target mesh might be too low quality to map to appropriately and might have a strangely different topography as well. How do you map a muzzle to a human face for instance? Or cat ears on top of the head?
Current real time facial recognition techniques find common points on a face and use those as points of reference. For instance it's very easy to find the corners of a person's mouth, corners and centers of their eyes, their chin, end of their nose, and the outsides of their face. This is because (barring mutation or mutilation) we all have the same general appearance. If you can find things like a person's mouth and chin, you can determine how a jaw bone should move. You can usually figure out where the cheeks are as well, and using the outside corners of the eyes and the corners of the mouth, figure out how the facial muscles are moving on a general level (bunching up for a smile, drooping for a scowl, squashing and stretching for a :\ face, etc.). These motions can all be translated to bones which are pre-rigged to move certain vertices very easily on the face. At least, very easily compared to figuring out how to move all the vertices of a face, compared to an entirely alien face (like a human to a dragon).
The reason I am asking this in the first place is when I see things like this : http://grico316.deviantart.com/art/.....tion-167815941
I consider the extra work learning how to do it a worthwhile payoff.
I have been 3d modeling and texturing for 8 years so I know enough about at least the basics of everything else but I am still trying to decide on a work-flow for facial rigging and both have their pros and cons, like you said, but I think I will go with bones out of personal preference. Even if it may be harder to learn.
I consider the extra work learning how to do it a worthwhile payoff.
I have been 3d modeling and texturing for 8 years so I know enough about at least the basics of everything else but I am still trying to decide on a work-flow for facial rigging and both have their pros and cons, like you said, but I think I will go with bones out of personal preference. Even if it may be harder to learn.
Honestly that arrangement is probably overkill for something like FaceRig. It's trying to simulate both the bones and the muscles over top of them (see the ring of bones making up the lips). In actuality you really need maybe half the bones in that head or less to do a good job, unless you're trying to do cinematic work. In which case most cinematic companies would be asking why you aren't using shape deforms. >D
There's two really big reasons to go with shape deforming over bones:
!. The expressions can look exactly the way you want them to, or a blend of expressions can be put together
2. Doing a phoneme chart with the mouth, chin, and cheeks all in the right places is dead easy, and you can very seamlessly transition between each of the poses to lipsync.
Bones can do lipsyncing as well, but not nearly as easily. They do allow you the same range of control though if you set them up right.
There's two really big reasons to go with shape deforming over bones:
!. The expressions can look exactly the way you want them to, or a blend of expressions can be put together
2. Doing a phoneme chart with the mouth, chin, and cheeks all in the right places is dead easy, and you can very seamlessly transition between each of the poses to lipsync.
Bones can do lipsyncing as well, but not nearly as easily. They do allow you the same range of control though if you set them up right.
Yeah, I am half asleep right now and when I typed my original response. Let me see if I can get this right on the second time. I am hoping that face rig becomes advanced enough to capture and record for rigs/blend shapes of cinematic complexity and in the end, wondering which work flow would work best for me in my personal projects.
There, I think that came out clear enough, maybe, probably not, need moar coffee...
There, I think that came out clear enough, maybe, probably not, need moar coffee...
Bones are, for most purposes, all you need. Most people agree with this until you start getting into cinematic work where your flaws are displayed for all to see on a screen larger than 6mx15m. If we're talking about your standard 18-24" computer monitor, especially about having your "face" taking up maybe 1/10th to 1/16th of a screen at best, those flaws are virtually non-existent if you did a good job with the bone rigging.
So wait, assuming you already have 3D software and 3D CG experience and all that is it possible to make your own working face for this on your own or not? As in, without any official involvement from the actual facerig devs. I've been looking forward to being able to make and sell customs faces since I first heard of facerig, but it seems like I keep finding conflicting information.
Verdammt, jetzt werd ich dich irgendwann einfach fragen MÜSSEN ob du commissions für Facerig Avatare annehmen wirst und wie teuer die werden... müsste ja irgendwann möglich sein eigene Avatare zu integrieren, oder?
Ja, uhm, auf jeden Fall: seeehr genial was du machst, nicht nur dieser hier, sondern alle deine Kreationen
Ja, uhm, auf jeden Fall: seeehr genial was du machst, nicht nur dieser hier, sondern alle deine Kreationen
Comments