For receiving Facial Mocap data, which is best: blendshapes or bones?

Kyle_butler

Hi, guys! I've searched the forum and found lengthy discussions about this topic. However, it didn't answer a specific question.
I'm planning to do my animation from mocap data, using a software like faceshift or Mycap.
By doing so, which is best to receive the pointcloud data, facial rigging with joints or blendshapes?
Thanks!

Bellsey

It's hard to give an absolute definitive answer as alot depend on the final output. Plus both have their pros and cons. Bones can give better optimization and lower memory footprint, but as bones are rotation data, you can loose detail if the head mesh is too low. You'll still need a decent amount of bones though to get the quality, 40-60 at least.

Blendshapes can be good because you can you're literally blending from one shape to another and you can ensure the detail is there, in the head meshes. But you'll need a good amount of blendshapes to ensure the output quality looks good.
Blendshapes is probably the (slightly) easiest to setup as many 3D pacakges have simple blendshape managers in them. However it's still better to create custom controllers to allow better interaction for animating.

Generally the way these facial capture systems work (video or optical capture), is that you create a facial rig in your respective 3D package, and then whichever tech you use, the data captured is retargeted onto that rig. Personally I use Softimage Face Robot for my facial stuff as frankly I've still yet to see anything as good and off the shelf. That uses an underlying bone rig and custom muscle solver, but I can export bones, blendshapes and point cache, if I wish to.

The best rigs I have seen and worked with, tend to be a mix of both bone and blendshapes

Kyle_butler

Bellsey, thanks for your answer. I'm modeling, texturing and rigging as if it was for a in game use, but in fact my project will be an animation in Maya, so I'm not too concerned about memory use.
I've been suggested to use Facerobot before, but by using it, this means that I would need to create a facerig inside Softimage, export the data, create a blendshaped head inside maya and then apply the data from facerobot?

Bellsey

You don't have to worry about creating a rig yourself in Face Robot, the setup process does it for you. If you aren't concerned about exporting to realtime, then your export options become more flexible as well.

Face Robot has a shape rig export that will help you create blendshapes and then export them to Maya. The export process will also automatically plug the blendshapes into Maya Blendshape editor and also export the basic rig so you can animate with it as well.
Or you could even geometry cache out from Face Robot and then simply attach it to you head mesh in Maya.

Kyle_butler

Bellsey, I had no idea that Face Robot took care of those stages. I'll give it a try a post the results here. Thank you!

Bellsey

You can find tutorials here: http://www.youtube.com/playlist?list=PL551AE1C29578EAE9

I did have a video of the actual shaperig export here on my YouTube channel: TheSuiteSpot
but it doesn't seem to be there. I will have a look around for it and upload if possible

Mark Dygert

I would build a system off of blend shapes/morph targets, they are easier to work with, easier to animate, easier to rig up to a control board, easier to capture poses from. After all is said and done you can always set up a bone rig to follow specific points on the blending face if you need a bone based rig at the end, even if the quality takes a bit of a hit at least you have a best case scenario to pull the animation from.

That way you can have an accurate capture for cinematic and a realtime bone based rig and you aren't handicapping yourself to only bones.

Kyle_butler

Bellsey, thank you for the link! Did you have any luck in finding the video you mentioned? Was this the "FaceRobot webinar" that some of your other videos refer to?

Mark, I'll probably go with the facerobot's generated blendshapes, for the same reasons you pointed out: easier to work, animate, etc.

Mark Dygert

Yea FaceRobot is awesome. I'm not a big fan of Softimage but for FaceRobot I deal with it. Good luck!

Bellsey

Kyle_butler wrote: »

Bellsey, thank you for the link! Did you have any luck in finding the video you mentioned? Was this the "FaceRobot webinar" that some of your other videos refer to?

Mark, I'll probably go with the facerobot's generated blendshapes, for the same reasons you pointed out: easier to work, animate, etc.

The webinar is here: http://www.autodesk.co.uk/adsk/servlet/pc/index?siteID=452932&id=17148448 under the title Facial Animaation worflow for games. This site isn't the best in the world as the recording shows the entire webinar window, but I don't have control of this setup. The webinar didn't show the blendshape process so I might try and find the video or maybe do a new one. You can still watch the webinar but I will try and upload the demo part to my Youtube channel.

Kyle_butler

Great, Bellsey. I've just watched the webinar, thanks for the link.
The youtube playlist you sent me will keep me busy for a while, but if you could upload the blendshape process, I'd really appreciate. I'll subscribe to your channel in case you find it.
By the way, have you heard about faceshift? I see great potential in it, but right now, the lipsync is not fullly developed.