SweetFluff R&D Report #3

I've been wandering in the depth of Blender hell (AKA learning Geometry Nodes) to search for a jiggle physics solution that I've been conceptualizing for a few years now. And I believe I'm actually quite close now to cracking the code for it.

 
Demonstrated on this non-human shaped mesh so it can be posted publicly, is a new in-house developed jiggle physics solution I've been working on that I'm calling Hyper Advanced Jiggle Physics.

The Problem with Traditional Solutions

The main problem that I see with traditional real-time jiggle physics (spring bones / wiggle bones) is that the torso orientation is not taken into account. When leaning forward, backwards, or to the side, traditional methods do not allow the breasts to droop or compress naturally relative to gravity.

On the flip-side, the issue when using nonreal-time, fully physics driven solutions, is that the resulting shapes post-calculation are not artist-friendly and are difficult to tweak. If a resulting shape from the simulation doesn't look correct, or if the artist desires to exaggerate or push the shape further, there is very little control the change or adjust it.

With Hyper Advance Jiggle Physics, I'm aiming for a hybrid approach that is both physics simulation driven and art directable.

How It Works

The basic mechanics of this tool is that the artist will manually sculpt multiple breast shapes for various torso orientations to their liking. A rigid body simulation (seen as the ball in the spherical cage beneath the mesh in this demo) is then run based on the movement of the character's armature. The rigid body simulation result is next processed through geometry nodes and outputted to those moving colored cubes. The height of those cubes then can then be read and influence which of the manually sculpted shapes should be currently active. This allows for very easy tuning for how the breast should look for every frame of the simulation.

However, this is not a wholistic solution and is currently not fully featured. This will need to be paired with additional solutions in order to account for things like collisions and secondary vertex-based jiggle motion, which I am currently experimenting with.

Vertex-based Jiggle Prototype

 I have also been developing this vertex-based jiggle geometry node tool that ideally would be used in conjunction with Hyper Advanced Jiggle Physics. This does not use any bones or pre-sculpted shapes, but just calculates each vertex's velocity over time relative to an empty, and applies inertia and dampening to that vertex's target location.

However, I still can't quite crack the math on this yet. As I can get it working only in a few small scenarios, but when I actually plug-in all the variables needed to work on a moving armatured character, the simulation just implodes and breaks down. So there's still quite a bit of work to put into this one.

This kind of secondary motion though would extremely boost the quality of my animations if it could be applied subtly across different areas of the body. So I will continue to look into creating the correct Geometry Nodes for this.

Conclusion

Though there is still quite a bit of work to put into the Hyper Advance Jiggle Physics solution, it is in a good enough place now where I can begin applying it to my Tifa animation model, and test-drive this new tool with some very short animations for her. So please stay tuned!