Statue smashing with DMM and Maya dynamics

Completed simulation shot with test render

Been given a big epic shot to do for the Cerebus film project. So have been pretty quiet on other things the last few weeks…

The shot requires Cerebus the Aardvark smashing and destroying an idol statue of himself in one continuous sequence with 3 main distinct action sequences. Smash Left leg, smash right leg, then main body of statue collapsing.

The animation was done by different animators with different world positions and no actual statue geometry to act to, so first painstaking task was to combine all the animation maya scenes into one cohesive character rig. Thankfully I was able to achieve this by using Maya’s trax editor to match, slice animation clips and using animation layers to offset global transforms to match each clips position. ( Big thanks for the riggers for using a cohesive naming system and using the character nodes in their rig, without it trax editor would not have worked as nicely )

First major hurdle out of the way, I then needed investigate how to shatter the statue, having used DMM previously for a smaller scene I knew it was capable of delivering good results. But with such a big object like the statue this was going to be hard.

Because the statue needed to be smashed with 3 different action sequences, I prepared the pre shattering in 3 distinct groups – R Leg, L Leg and torso/head. This gave me the ability to scale up or down the amount of shattered parts ( tets ) as needed.

By breaking the sim down to 3 parts I was also able to save simulation time, so good planning and RnD helped! Will post more details on the DMM part of the project on a later post, for now lets see what I came up with.

^ This is the final simulation from DMM roughly around 25,000 tets. It was tricky to get the shards to ‘act’, but its come out nicely.

There were concerns about the blockyness of the shards, so I addressed the concern by doing a test render with some very simple dust cards composited on top.


^ Test render with motion blur and displacement.

As you can see the final render looks pretty good once displacement and motionblur is added to the inner faces of the shards. And I needed to find a good compromise between speed and quality without blowing up my single machine. I’m sure if we were at ILM we could have millions of shards, unfortunately we are not :p

To give you an idea of the scale of simulation here. The cached DMM simulation is aprox 4.5gb worth of data. Took over night to simulate ( 10hrs ) at .5 frames over 800 frames. This does not include the fluid or particle sims to come later on.

The trick to good simulation work is to layer it up with different types of dynamics, so next up I’ll be tackling the dynamic particle side of the scene, which will really bring it to life. After some RnD I decided for now it would be best to use Maya fluids, and instanced particles for the additional 2 layers.

Layer 1 – Base simulation with main shattered parts of statue.
Layer 2 – Smaller rock sized particles emitted on smash. ( Emit instanced particles )
Layer 3 – Maya fluids to simulate the very fine dust particles and tendrils. ( emit particles from shards into maya fluid )

Here are some links I found useful as a reference.

RESCUE 3D – Cathedral shot
Emitting Maya Fluid From Dmm Object
Dmm And Maya Particles ( Emitting Maya Particles From Dmm Object )
Emitting Maya Particles From Dmm Fracturing faces
Waterfall using fluids and nParticles
Pixelux DMM Forums
VFX BREAKDOWN | PIRATE | 030_0140

2 Comments

  1. Hi John,

    Nice job on the Cerebrus fracture! For making the shards look better, you might try a trick of creating a super-high tesselated object using the DMM tools. Save out the resulting tetrahedral mesh as an OBJ and then use that as a splinter cage for a lower-res tet mesh. I have gotten really good results with that. The alternative would be to sim at a million tets and then decimate the resulting geometry faces to improve detail.

    Cheers,
    Vik

    Like

    1. Thanks for visiting my blog Vik! I’ve posted the final sim and test render video above. The problem I found with the old version of DMM was the initial shatter calculation when opening a scene, this hindered my ability to have more tets. For the playblasts and adding particles, I cached it out as geometry cache, but the problem is the textures do not work properly. So at render time I have to resort back to the cache DMM scene. I think this issue has been addressed in the new version of DMM, which I have yet to try out.

      Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s