A First look at Illuminate labs’ Turtle for Maya

A First look at Illuminate labs’ Turtle for Maya
By Christopher Nichols

Christopher Nichols is an ex-architect that was a visualization expert at a large architecture firm. He currently is working for Digital Domain in Venice California, and has worked on such films as “The Day After Tomorrow” and “I, Robot.” The views expressed in the article are those of the author and not necessarily those of CGarchitect.com or Digital Domain.

Prologue

Before I start to tell you about Illuminate labs’ Turtle rendering package, we will need a little perspective. In fact we will need two perspectives.

Lets start with some history based on the ArchViz community. Many of you have been around the world of Global Illumination for many years. Of all the subgroups in the CG artists, you have the most experience then any of them with GI. I remember my first try at lightscape when in grad school at around 10 years ago. However, in the last 3 to 4 years, raytracers are starting to take over the world of GI, and the average CG architect is in heaven… well at least the ones that use 3dsmax. It is one of the most common 3d platforms of the archviz community, and has allowed CG architects to be flooded with new, fast, and powerful rendering engines, such as Brazil, finalRender, Vray, and more. Discreet even provided several very solid methods of doing GI right in 3dsmax native package, including the radiosity engine from lightscape, and a free copy of MentalRay. Not to say that any of the native max tools, or any of the third party tools are perfect. The point I am trying to make is that the archviz people have two main things when it comes to GI rendering: experience and choice.




Now imagine the VFX community, a smaller (in numbers) community, but with a much larger reputation. They are generally the trendsetters in the CG world. Here is the irony: they have not had the privilege, or luxury of using GI everyday as you have. In fact, they have been struggling to find a way to use it. If they do, it tends to be a lot more limited. Why is this? It is generally for the reason that they have a very set pipeline. While 3dsmax, and several of their third party rendering engines make it in to the VFX world every now and then, they are generally a minority, and a fairly small one at that. While this group may be growing, and are now immersed in the world of GI, the real powerhouse of the VFX world is Alias’Maya, and Pixar’s Renderman. I will not really get into why this is the case, but you can rest assured, that a very large majority of all the big special effects shots that you might have seen on the screen last summer (and every summer in the last 10 years), have been done with this combination. This is the group of people that Illuminatelabs’ Turtle is marketing to.

Introduction

First, lets look at some of the important features Turtle offers: Global Illumination, Caustics, Final Gathering, Ambient Occlusion, Image Based Lighting, Sub-Surface Scattering, Motion Blur (2D and 3D), Depth of Field effects, Texture baking, and some cool geometry tools, such as SubD and Nurbs rendering, and Displacement rendering. As you can see, it seems to do it all.

One thing I could not seem to find, especially if they are trying to market to the VFX community, is the ability to write custom shaders. This would be a critical option.

The simplest way to get an idea of what a rendering engine can do is to do tutorials. This is assuming that the tutorials are good. I found them to be simplistic. But really this was not a problem. Reason being, Turtle seems very easy to learn. If you have done any type of GI rendering before, you can pick up nearly all of Turtle in about an hour or two. So just browsing the tutorial files was really enough to know my way around the interface and features.

During the time that I was testing Turtle, an updated beta version was released which the makers of Turtle recommended I try out. Several improvements were made to the new updated version, including the addition of multi-threading. I was actually a little shocked when I first tried Turtle that it did not have multi-threading. Other key improvements happened along the way, including some large performance and speed impovements. All of these are signs that Turtle seems to be growing quickly, as many of the young rendering engines experience in their first year.

Lets look at some specifics

Global Illumination

Turtle does its GI mainly via photon mapping, with the option to use finalGather for the “finesse” pass. If you have ever used MentalRay, it is almost identical in its process. Turtle has two methods of doing photon mapping: Standard Photon Mapping (similar to mentalRay), and “Liquid Light” photon mapping. The Liquid Light method is pretty cool, fast, and “looks” different. Besides the fact that it appears to be at least twice as fast, visually it is easier to “dial in” compared to standard photon mapping since it breaks things down into a grid rather than those ugly photon mapping splotches. According to the manual the Standard method is more accurate than the Liquid Light method. With that said, like with most photon mapping methods, you can’t really count on photon mapping alone. The results have way to many artifacts. For this reason, you need to add final gathering. This is similar to any other package that you may have used. It refines the results, gets the GI detail in all the corners, and tends to smooth out the artifacts.

Exterior rendering with area shadow, and final gather skylight.
Rendering time 8min17


Interior rendering torture test. 20 bounces of light, liquid light photon mapping, final gather. Rendering time 4min11

 

Overall I found that Turtle did fairly well in the GI tests that I did. I will have to say that at first I was not very impressed, but there seems to have been a great deal of optimizations in the new beta version which make it very hopeful. Some of the earlier tests which seemed very slow, seemed to speed up by a factor of 5 or 10. I will restrict myself from giving specific speed comparisons. Anyone who has played with several raytracers knows that it is easy to skew the results either way by accidentally tweaking some settings. With that said, I would say that it is not the fastest GI rendering engine I have seen, but it is fast, and is probably the fastest Maya one I have I used. But based on my prologue, you already know that the competition is fairly slim in Maya these days.

 

Image based lighting

As anyone that has used a GI rendering engine would know, the way that most people would do Image based lighting is to use a “skylight” method with a HDRI in the environment lighting. Normally, this would be done in a stage equivalent to the final gather stage. Ok, remember the time you were poor and could not afford a fast raytracer? There used to be these tricks to try to get that GI look without having a GI rendering engine. It usually involved creating a dome of spotlights that would get their sample value from the HDRI file. Similar to this link here: http://www.3dluvr.com/subagio/domelight/instructions.html maxscript. Maya had the same thing, called here: http://www.3dluvr.com/pixho/toolbox/mel/GI_Joe/GI_Joe_doc.html. I was shocked to find out that turtle would consider this a serious method of doing image-based lighting. This method is filled with problems, especially for architecture models. On top of which, it also takes a long time for turtle to be able to sample the HDRI file. Image based lighting is actually a critical part of GI, and is very important for archviz. The way that Turtle decided to implement it was shocking and frankly, completely useless for archviz.
 

Image based lighting with final gather skylight of HDRI


Image based light feature

But wait! All is not lost! Hidden in the documents, you will see that you have the ability to attach an HDRI environment file to the final gather skylight. Doing this will give you the more predictable Image based lighting, or at least the type of image based lighting that you are used to. In fact, Turtle does this very well. I would recommend that if you would like to do image based lighting, use this method and not the “feature” that is included in the render.

Ambient Occlusion

This section needs a bit of back history, and a warning. First the warning: I have a strong personal opinion regarding Ambient Occlusion, which may present itself in this section. Some of my opinions are debatable. Now the history: Remember when I said that the VFX industry does not always have the luxury of full GI? Here is a link on what it a more common solution:

http://www-viz.tamu.edu/students/bmoyer/617/ambocc/

Here is a link written by Andrew Whitehurst, which outlines the steps to doing an Ambient Occlusion pass. Note that this is before renderman had Raytracing, hence the use of a dome of lights:

http://www.andrew-whitehurst.net/amb_occlude.html

Here is my basic explanation. All it is, is a Skylight lighting, where the skylight is 100% white, with no bounces of light. All it does, is give you the contact shadows and a darkening of the creases. What you then can do is to take an environment image that you might use for a reflection pass (or image based lighting). Then blur it a LOT, until it is simply a blob of color. Then do a reflection pass with that. Don’t raytrace it, just a simple environment reflection will do. Then you can multiply the Ambient Occlusion pass with the environment pass in a post process (either in the rendering or in a compositing program), and you have the simple hack for doing something that starts to approach Image Based Lighting. It allowed people to make lighting look very diffuse. This method was outlines at Siggraph 2002, and was used on Pearl Harbor, which was in production around 1999. While this method is still used to help in diffuse lighting. It is mainly used because they do not have the luxury of a fast Raytracer. Anyone who did have access to a fast raytracer such as Brazil, Vray, or finalRender, would simply do REAL image based lighting. The same thing would be true for Turtle since it seems to have a fast engine.

The method that Turtle decided to implement is identical to the methods outlined above. On the other hand, if one was to use Ambient Occlusion in their process, especially for an animation, they did include a way of baking the AO into the model through Texture Baking. They also give you the ability to adjust ray distance, cone angles, and other parameters that are commonly used in Ambient Occlusion shaders.

Again let me explain that many people would say that Ambient Occlusion is a perfectly valid way of creating a GI look, and those people will be happy to have AO built in. I would say that in time, people will realize that GI is beyond soft contact shadows, and they will want light to bounce around and give off some color bleeding. Also keep in mind that Ambient Occlusion does not work for interior lighting since it is a primitive skydome that does not bounce rays, so there is little benefit for architects. The developers of Turtle however did not intend for this feature to be used for architectual visualization, but rather game developers.

Anti-Aliasing

When I first examined the Anti-Aliasing option in Turtle, I found them to be Sub-par. I was pleased to see that they had made some critical changes in the beta version which improved its quality a great deal. It offers an adaptive sample based anti-aliasing method, which several filtering methods, including all the usually suspects (box, gaussian, Catmull-Rom, Mitchel, etc…). My main complaint about the current AA, is that one needs to increase the samples a great deal to get a good enough result.


Production AA at default values. Rendering in 4 seconds


Production AA with values turned up for better results. Rendering in 20 seconds

Editor's Note: Illuminate Labs is aware of the slow AA render times and will be adressing this in the next release.

 

Displacement

Turtle also offers displacement and SubD rendering. These sort of go hand in hand. These are important features that need to be in any modern rendering engine. I was happy to see that turtle had this feature. I have a few issues with it, which I hope will improve in future releases. For example, while it is possible to use procedural shaders for the displacement, they need to be prerendered as a texture first. This is not always a good idea for things like landscapes. Also, other rendering engines such as Renderman and Vray offer more optimal displacement by making it view-dependent, which turtle does not have.

Example of the use of Subdivided surfaces, and displacement mapping

 

Conclusion

You will remember that in my prologue, I mentioned that the archviz subgroup of CG artists have the most experience in GI then any other group. So while I was not completely blown away by this product, I was pretty happy with what it could do. For the poor Maya people that don’t have luxury of what other 3dsmax users have had for the last 3 years, welcome to a new world. Turtle is still young, and still has a lot of kinks that may need to be worked out, but having seen the changes from their first version that I tested to their latest beta, I can see that they are going on the right track. And for those of you that are using Maya for archviz (apparently very few according to this link here: http://www.cgarchitect.com/vb/showthread.php?t=7757, Turtle may be the first step that you need to start to get the look that many other have been getting. Another great advantage of Turtle is that it works all forms of Maya: Windows, Linux, and MacOS.

There is one thing to keep in mind. Turtle seems to be priced similarly to Mental Ray and Renderman. At $1,199 per workstation, and $999 per render node, it is pretty much pricing itself out of the archViz group. Considering the fact that Brazil, Vray, and even MentalRay is cheaper then that, I find that Turtle may be overpriced.

For more informaton on Turtle, you can visit the Illuminate Labs Website here: http://www.illuminatelabs.com/

 

If you would like to post comments or questions about this review, please visit our forum