End User Event or EUE, is famously the "only industry conference that takes place in a pub". This year, I will be speaking at EUE in Utretch, Netherlands, about our 8 year research in terrain design which culminates into The TOR Platform.
...believability is more important than realism because what we often create is not real or even realistic.
Where it all began
After a successful jungle exploration across Costa Rica in 2008, Cynthia and I set out on expedition into the US Southwest the following year. Over 4 weeks we traversed Arizona, Utah, Colorado, and New Mexico, taking a fresh look at terrain design. This time with an artist's view instead of a programmer's.
Adhering to an artistic perspective is paramount. Our tools are made to help artists achieve their creative vision. And especially for environment artists, believability becomes more important than realism, because what we often create is not real or even realistic. But it has to be believable to transport you - the viewer - to the fictional world the artist has conceived.
With this in mind, we looked at "unbelievable" formations in the Southwest - strange erosive patterns, inverted voronoi pockmarks, and massively carved out structures that seemed to defy gravity. We started listing what aspects contributed to "realism" and which aspects contributed to "believability". This began our long journey into terrain design reform.
The TOR Platform
The rendering power to bring virtual worlds to life has increased dramatically in the last decade. However, the tools to create procedural environments has not kept up, and in some areas, has been stagnant.
The above is a qoute from the EUE talk description.
In general, the industry has been treating procedural terrain design too much like making materials. If you are a seasoned noise-tinkerer, there is nothing wrong with this approach. But too many people don't have the knowledge or capacity to delve deep into how procedural noises truly work and how they need to be adapted to terrain design. One major consequence is unduly heavy erosion to attempt to make noises look believable. This in turn can cause CG terrains to become obvious and predictable.
With GeoGlyph, we started encapsulating goal oriented design into individual components. This made it easier for artists to achieve shapes based on type, rather than trying to wrangle a noise into a specific form. The TOR Platform is the natural evolution of GeoGlyph.
Dissolving the shader approach
Procedural design has always been dominated by noises. But for terrain design, this has meant stagnation. Perlin and Voronoi/Worley noise is still the root of all terrains, limiting artists in the same create-filter-erode chain of shader-style design.
QuadSpinner has spent the last few years comparing hundreds of terrain graphs to identify the patterns of how artists created terrains - especially professionals who need to get good results quickly.
The TOR Platform changes the existing paradigm of terrain design by dissolving the shader style approach of noises and filters, and augmenting it with elements closer to their real world counterparts.
In simplified terms, this means a valley should be an actual valley object and not a shape we create then subtract out of the terrain. It means, that the artist should have control over erosion and not be a slave to it. If you want your river lines to be stronger, it should be an artistic decision and not up to the limits of the erosion algorithm. It means, we need a better balance between what we achieve through procedural means and how much creative control we have over these procedural algorithms.
Here is a basic example: Perlin Noise is ubiquitous in all procedural design applications. However, using Perlin noise, or its famous cousin Voronoi/Worley, requires a fair bit of modification before it can truly look like a terrain.
With the TOR Platform, we altered the noises so that they look like terrains out of the box. Our goal is to make each and every component look and feel like part of a terrain, even without erosion.
Fall in love with erosion all over again
Hydraulic Erosion has stagnated over the years. While the original concept Ken Musgrave put forward still creates the basis for all hydraulic erosion, we believe it is time to componentize this type of erosion to give artists deeper control. In the TOR Platform, we wrote new forms of hydraulic erosion that separate the sedimentary and channel-carving processes.
In this image, the wide, sedimentary flows are separately applied from the deep channeling and river generation. Not only do you get detailed control over both processes, a large part of it is performed as a post-process. This means making small alterations or exploring design options does not require you to rebuild your terrain over and over again.
TOR differentiates between simple sediments such as sand and soil, and larger debris such as scree and individual rocks.
Some of the erosion nodes in TOR are: Sediments, Debris, DeepChannel, and Breaker. They each comprise an individual aspect of traditional Hydraulic Erosion. The Simple Erosion node provides a quick way to control all these separate processes when detailed control is not required.
TOR's Erosion Studio gives the artist direct control over hydraulic erosion. Instead of sliders and numbers, you pick from a variety of erosive tools like you would in a paint program, and then assume total control over the erosion.
This form of non-linear, non-uniform erosion mimics the real world far more than the regular, uniform erosion we are all too familiar with. When two different erosive flows collide, they affect each other and their surroundings in a natural fashion where physics plays a greater role.
A detailed overview of Erosion Studio will be presented at EUE, covering features such as Infinite Flow, Rock Deposits, and animated erosion output.
All of the above is just the tip of the iceberg (or 0.99f of the heightfield, if you prefer).
The talk is entitled Reinvigorating large scale digital environments and will take place at End User Event 2017. If you are attending, please drop by - I'd love to say hello.