Why though?
I was playing God of War Ragnarök during Christmas break, and I really liked how the mystic gateway animation felt. During one of the play sessions, the developer side of me woke up and thought, "Hmm, I wonder if they just reversed an RBD sim for this."
Like any technical artist out there, we like to tinker, so I thought, "Let's try it out! I'm curious how this was made." And I did. I created a fast prototype in an hour or two, proving that reversing a gravity RBD sim would be enough. And that was it. I moved on to the next thing that caught my attention.
One month later, I stumbled upon the reference photo I took for this test and decided I could spend a bit more time on this and maybe write a blog about my approach to it.
While this won't be anything fancy or uncommon, I think it's interesting to see how others approach problems and how it can sometimes help with other unrelated things.
Before I ramble onwards, two disclaimers:
- This past year I've been learning Houdini, so my breakdown will focus on a Houdini workflow(although similar concepts exist in other DCCs as well).
- This is a fast prototype, where i don't care about polish or quality, its just to prove a concept.
First step, references!
.gif?1707942861)
Break down part
Looking at the images I noticed:
- There are 2 states (Active and Inactive), we can also see that there's an archway shape and some rocks following this shape.
- Rocks are following the normal of the curve towards the inside and bend with the curvature.
- There are gaps between the rocks and the wood frame and that makes our life a bit easier to recreate this.
This gives me plenty of information on how I can start, but lets also look at the animation and see what our model needs to support in terms of movement.
- There are 2 big rock pillars intact and not moving (left and right, one a bit uneven)
- Each individual plank fracture independently, indicating it's probably split into its own element.
To-Do List:
I need a rock asset, an archway made with it, a wooden door made of vertical planks, a fracture for the door and an RBD sim.
Rock
This was easy, and I cheated a bit, I downloaded a rock from Megascans. While Houdini can easily generate some, I thought this is not the real focus here.
Archway
For the archway, I created a spline that i mirrored and remeshed, where a copy to points was used to duplicate and scatter the rock asset. To make it look less repetitive I'm adding some random scale and a random rotation, while still trying to maintain the rocks rotation towards the curve interior.
Wooden Door
Since I knew I would need to fracture the wooden door, I wanted to have separate planks created from the start, so I don't have issues with newly created faces. For plank creation, I created a shape from the archway initial curves, bridged the missing gap, and using a boolean from the door shape with boxes created as planks it results in planks following the archway shape.
Wood Fracturing
Fracturing the wood planks, I went with the old approach of scaling down the mesh into a Voronoi fracture and then scaling it back. This workflow creates some wood-like splinter fragments. Afterward, I handled the cluster creation manually by grouping up splinters that are close by.
Simulation
Tearing down using a RBD Bullet solver, just an RBD Solver with gravity on.
And now final step, reversing the simulation of the RBD solver.
This was a fun thing to tackle, especially since I didn't have to worry about the final quality/polish of it being in a real production environment.
I want to emphasize that having a quick prototype does not actually represent the entire amount of work needed for an asset to be used in a shipped title. It's more than just having a working animation; it's about visual, gameplay polish and production iteration. It needs to work with gameplay, collision, metrics, combat, VFX, and audio setups. It needs to work in a full game environment and, more importantly, be iteration-friendly.
I always like to find things that I could have done differently or improve on, and for this mini-project, it's no different. This is what I would have done differently or had a different approach if it were in a real production environment:
- Rock generator that adds that visual noise as repetition is still pretty noticeable in my scene(or more rock assets to scatter) even if i tried to hide it with rotation/random scale etc.
- I would like to find a nice way to check intersection between the rock points, as the current setup i was just lucky with the seeds and scale of rocks that i used, my current setup was NOT production friendly, and would have required more work if an art director or shot supervisor would have requested lets say...add 1 more rock. *flip table*
What are your thoughts on this approach?
/Andrei
P.S: Adding also the Houdini network view.
