Vertex perfect weight painting

Just asking if anyone has an approach on precise, Vertex perfect weight painting?
That is sure and confident to breakdown parts and muscles into understandable pieces, and spreading weights evenly around?

One probably has their own unique approach, but i am particularly looking for the kind of workflow that goes into games like kingdom hearts and final fantasy since yonder times… i wonder if they use particular method to paint vertex weights and a way to think precisely so that the animation can be smooth and nice?


Keep in mind that “smooth and nice” animation depends as much on good topology as it does on precise vertex weighting. And on the animator, of course!

With that in mind, the most practical workflow regardless of the use for the models is from broad to fine, starting with weights applied generally, without a great deal of precision – automatic weighting is an example. Deformations are then tested and the weights adjusted as needed, making a number of passes with different poses that push the models to fair extremes, with particular attention paid to troublesome joints like the shoulders. In Blender the Weight Painting tool can accomplish these steps, down to the level of individual vertices if need be. If the model has special requirements that require a further round of fine tuning (my muscle system is an example of this), then the individual vertex weights accessible in the 3D Window>Properties panel (N-key) can be tweaked to three decimal points as needed.

I can’t imagine that any workflow involving vertex weighting for armature influence would differ much from this basic outline, as the principles apply no matter the specific application.

Hey, great idea! I don’t have any particular comment about your muscle system, other than they look impressive…
But i guess i’m looking for something finer, since the polygon count runs on less than 2500 for each character, and every single parts
from head, hair, apparels, legs and arms, probably each take 500 and less polygon counts.

So when it is broken down… I only have so much polygon in that i need to paint…

And even though it is “not that much” compared to many other people, i’m wondering
how can i decide on the get go, how much weight paint should go to particular vertices, and its particular motion?

Because reading your comment means that it seems much better to record whole actions first,from idles to special moves, and then fine tuning weights later,
given the actions are all set from the design guideline, and then the weight painting can follow…
And although these are not bad, given the animation is also another headache to be concerned…
Do you think there’s a way to address these in a much more efficient, on paper precision?

Learning to do this comes from experience, there is not a formula or method to replace that. That’s why automatic weighting, while not very accurate in many cases, is still a viable initial step, as it provides a basic vertex weights structure that can be adjusted as needed by the actual use of the model. The Auto Weighting algorithms are designed to quickly produce a reasonably useful starting point; after taking a few dozen models to finished vertex weighting, one can learn enough about the process to dispense with it, but it is still faster than applying weights manually to each joint from the get-go, as you say.

I think you misunderstand my comment – recording Actions before the weighting is complete is not a good idea at all. However, to properly analyze how well the vertex weight distribution is working, the model must be deformed through poses, and the more extreme the pose, the greater the challenge to the accuracy of the weight painting. Poses are not the same as Actions, though using the key poses of a character’s design sheet is a good starting point. Poses are also static, and Actions are dynamic, so even after the vertex weight editing the poses require it’s a good idea to try out some basic animation moves to see how the mesh flows from one deformation to the next. Shoulders are particularly sneaky in this regard, requiring a lot of attention to detailed vertex weighting and often some compromises based on the pose and/or Action requirements – rarely does a set of weights on a particular joint perfectly satisfy all possible deformations.

Took me quite awhile to break this comment down… yes, recording actions first before ANY weighting is probably terrible.
But it just strikes me rather stressful to test pose to pose when the action itself may involve extreme angles.

Like, how far do one test until the weighting feels good for any possible circumstances?
Not taking games like SIMS or things like PIxar which tend to involve really realistic approach and step by step workload…
But more extreme actions and maybe a very narratively important “exagerattion” expressions, as disney’s 12 rule of animation says…
Can only be achieved by, solving all the possible actions first right?

My question probably is… since arm and leg has about 360 degree of rotation + scapulaes, necks and so on, plus
clothes and hairs that also uses bones in games and some instances of filmmaking…
How much should one paint and test until it’s done, in favour of efficiency and time delivery?
Rather than, splitting the weight painting into say 3 different values range from 0 - 1.000, making it a much more mechanical
but efficient work…
versus, testing all the motion out (with prelim weighting) and working out the weighting with all the motion solved?

All these ideas are all done in favour of, technical efficiency and product delivery…

Each joint has its own particular set of flex limitations, and when I’m doing vertex weighting, I usually flex them to their maxima as part of the testing. Your comment about certain joints having 360 deg. of rotation is not at all true – every joint has inherent limitations that should be taken into account. For example, bending knees and elbows “backwards” to test deformation isn’t useful unless your character is Plastic Man (j/k), and even with shoulders, the joint with the most degrees of freedom, rotation along any one axis has limits. So, bend a joint in its natural direction as far as it can practically go, inspect the deformations, and adjust the vertex weights as needed. This is what I mean by “poses,” not necessarily something the character would do as part of the “acting,” but instead a means of examining the mesh response in extreme poses, purely a mechanical consideration. Even if the character may never reach these extremes, doing so will insure that lesser flexing is also correctly weighted.

This process will give you a good foundation for setting actual keyframe poses during animation planning. If any deformation problems show up (and they are bound to, no first pass is ever perfect), then the weighting is adjusted again as necessary. It’s not always a single-pass solution. Production schedules may limit the number of corrective adjustments can be made, but calling the job done without making some tests of both keyframe poses and full animation sequences is not a good idea, there are always situations not anticipated or changes made to animations that were not accounted for in the first pass at vertex weighting.