Houdini 17 just revealed a new Terrain tool, with an improved Erosion system, much more powerful than the one we had on Houdini 16.5. When we had to use software like WorldMachine to get a realistic erosion effect on terrain, with a nice representation of the impact of water on the ground, it seems Houdini now is totally up to date, which is awesome as it’s always easier to work with a software in the pipeline, connected to our Database, Libraries of assets, etc… The Scattering has also been improved a lot, with now a MultiScattering tool with the beginning of what could be an Ecosystem simulator.
You can watch the presentation here, at 53′
Working with Ecosystems, rather than just scattering assets, in case of vegetation scattering of course, is probably the next big challenge we’ll have to face.
Today, when we want to scatter some vegetation on a terrain, we mainly rely on procedural scattering using a bunch of masks and basic rules, like height, slope, painted masks, etc… But some company started to invest into the development of a proper Ecosystem Software. Weta for Apes 3 developed their own Ecosystem simulator called Totara. Here is what the supervisor said about it :
You also mentioned an advance in growing the environment.
That was one of the biggest steps forward. We wrote a new ecosystem simulator that we call Totara after a native New Zealand tree. Our older Lumberjack program was great for growing a single tree, but what makes a jungle or a forest realistic isn’t just one tree, but how the trees relate to one another. One tree’s shadow may influence another tree to push higher. The limbs on the lower part die off and fall away. Shade-tolerant plants thrive in the understories. These things are very laborious for a human to go in and build and prune manually, and it’s a challenge to get them to look right.
How does Totara grow trees?
We set up growth rules for species of plants and for what environment each species thrives in. Then we throw a bunch of seeds together, a mix of species, and distribute them across, in this case, a mountain terrain where there is a concentration of resources. As the seeds start to grow, they compete with each other for sunlight and nutrients. They orient to the sun path. Some grow faster. Some can take colder temperatures and can grow at higher altitudes. We run the simulation over 80 years or more. As the trees grow, they influence each other. The distribution and arrangement of trees and pockets emerge naturally as the trees influence each other. As the altitude rises, the species change. We get correct negative spaces and undergrowth where branches won’t grow.
Can you art-direct the simulation?
We can freeze the simulation, shift trees, and make adjustments. We grow it in patches — grow one mountain and then another. The simulation tends to finish overnight. The simulations are pretty manageable.
Did you have other simulations?
We had a fair few. We had to problem-solve the waterfall — making it feel right in volume, scale, and participation between air and water. We wanted to make it feel like all the elements were coupled — the water particles drag elements in the air volume and the air affects the mist. We did the same thing with explosions.
Because of the run-up — we usually have to let a simulation run for as many frames as are in the shots — and because of the long shots, rather than overnight simulations, the big waterfall shots could take two days.
You can also see it in action in this video at 2:40
So why spending resources to develop a proper Ecosystem simulator instead of just scattering on points a bit randomly ? The main reason is that if scattering assets garanties the fact they will all be in the same lighting, will be in the right perspective, and have a good scale, it doesn’t mean that the way they will be scattered is going to look natural. It’s a lot of work to technically be able to achieve something like a complex scatter of trees, plants, rocks, adding some water, having a physical sky, god rays, etc… but unfortunately, it’s not enough…
The Ecosystem will take care of the influence of the terrain on the growth, the influence of a species on another one, how plants and tree will grow to catch more light, how they will depend on the shadows, etc… It is a real simulator, you can also see how a forest would look in 200 years, by applying real rules of life. It’s not the same process. We are not reproducing something, we are creating life.
If we are in the case of a standard scatterer, without Ecosystem, at least, before starting this sort of task, the Env Artist absolutely needs to look for photographic reference, something to match, which will show him how nature would have spread the trees, which sort of variations he will need, variations in shape, colors, size, orientation, etc…
If you look at the video from RaynaultVFX in another post, about scattering trees in Clarisse, it shows that one scatter is only dedicated to accident, trees without leaves, with a weird shape, dead trees, etc… And that step is crucial when you want to elaborate an environment based on scatter.
The most difficult is going to be to select the asset to scatter, the amount of variant, and to find a natural way to scatter them. Nature is way more interesting that an algorithm based on a noise or few masks, because it takes in consideration the winds, the slopes, the terrain itself.
Clarisse, Vue, Terragen, or Houdini offers amazing scatterer, but developing a proper Ecosystem, with rules based on nature, is probably the next step ! And actually, it seems totally doable ! So let’s do it !