Questions about facial rigging and blendshapes creation

#1

Hi there,

I´m coming from Max and even there rigging isn´t my specialty, but now I have to create blendshapes for the main character of the short film I´m working on for our rigger.
Last time I created a facial rig, I worked off from a tutorial by Paul Neale, which used morph targets (aka max blendshapes) to create all the expressions.
In this tutorial it was possible to just model symetrical shapes and then select the influences in the morpher to create the asymetrical expressions.
There were some issues with this approach concerning the middle vertices specifically for the mouth shapes.
because separating them into left right shapes and then combining them back together made them add up.

So I was wondering if someone could point me in the direction of how to approach this for maya? how do I properly model blendshapes for left and right separately? What do I need to care about (apart from the obvious to separate the eye area from the mouth area for example)?
I use Zbrush for the modelling part and then export FBX for each blendshape.

#2

Hi
You could use this Tool . Also check this where it point to some useful stuff for sculpting.
Regards,

#3

Thanks, but thats not quite what I was looking for.
Like I said, I don´t use maya, I use Zbrush for the modelling and I just wanna know what guidelines to follow for creating blendshapes, so I don´t get weird additive shapes when combining left and right sides for example.

Or how to deal with how the opening jaw (controlled by a joint) reacts with several blendshapes.

I´m guessing there is no easy answer here…but I already found this as a good starting point:

https://imotions.com/blog/facial-action-coding-system/

But also the sheer numbers of blendshapes for a character like gollum points to a more complex problem here…

#4

The book Stop Staring by Jason Osipia covers this for Maya. Not sure if there’s a newer approach to this using the blendshape editor. I don’t think so.

In Maya you use the paint blend shapes weight tool to separate the two sides. I’m rusty on the details but you paint out half the weight then duplicate the mesh to save it as a new shape. Then you need to invert the painted weight and apply it to make the other side. I think the book comes with some scripts to help the process but nothing that can’t be done manually. To invert the weight you can export the weight map and invert it in Photoshop.

In the book he also uses this technique to separate the brows and mouth into several parts. By sculpting a single large scale shape like “mouth up” or “brow up” and then splitting it into several parts using paint weights tool, you are guaranteed the parts will work together when re-combined.

Hope that makes sense.

#5

Well, like I said, I don´t use maya to create the blendshapes, but thats basically the same approach I used in Zbrush.

  1. Store morphtarget for neutral expression.
  2. Sculpt symetrical blendshape (for a smiling mouth for example) on extra sculpt layer.
  3. Duplicate layer twice.
  4. On the duplicated layer, mask out one half of the face.
  5. Use moprh brush to clear out the unmasked area.
  6. Repeat for the other side.

In this workflow I have to be careful with three things:

  1. I have to makes sure I clean out everything from the unmasked area.
  2. I have to try and not change the model in the middle. For example: if I pull up the mouth in the middle for the smile and then clean out one half, it will have a weird asymetrical smile.
    To work out around that issue, I can soften the falloff to the middle. This will result in the left and right side not adding up to the originally sculpted symetrical shape.
  3. For any necessary corrections like this I can then add another layer, where I just add the corrections to get to the original shape (leave the left and right side enabled while sculpting the corrections on a new layer).

I think thats probably the basic workflow, its just easier doing this in Zbrush than in Max, where you can see results as you model them and don´t have to predict whats gonna happen in advance…

The FACS Coding system is still super helpful…

#6

In the workflow I describe the original blendshapes are sculpted in Zbrush and then in Maya the paint weights tool is used to split them into parts. It sounds like your all zbrush method is very similar. You still might get a pointer or two from the Osipia book.

Good luck with your project.

#7

FWIW, this will do a similar thing in Maya. It splits a blend shape into two halves, with a soft blend region in the middle. It doesn’t do anything fancier like splitting based on masks for different parts of the face (that would be interesting…).