This is a follow up article to the one I posted most recently on how to light clear glass on a light background. I will refer back to that article, so it may be helpful to look at it before you read this one.

I started this exercise from my white field file. My initial thought was: just invert the background color to black, and put a white card where the black cards were. Problem solved! However, the solution was a little more complicated.

The key differences between the light field lighting scene and this dark field scene are:

1) Narrower field of view to cull out unwanted reflections.

2) Less of the contrasting background surface area needed to create edge reflections.

3) The background is further away to minimize key light contributing to background illumination.

4) The in-camera background is very dark, instead of light.

I used a night time interior level of exposure control (2 EV), which I should have used in both studio scenes. It’s important to set the exposure property, so that your lighting will behave realistically when global illumination is calculated. In these particular scenes, almost all of the reflections are direct instead of diffuse indirect, so it’s less important to match exposure and light values. However, it’s a good practice to specify real-world light and exposure values, in the same way it’s important to model at real-world scale -Your renderings will be more predictable.

I wanted a narrower field of view to control the direction of the reflections on the bottles, so I lengthened the lens. In order to keep the same image size of the bottles in the frame, I moved the camera back away from the scene. Because of this, I had to make the “table” longer, so it would fill the view without a gap at the bottom of the frame.

Notice how the background (bg) light and the teapot special light don’t cast shadows. This is a rendering optimization, since shadows are not needed on those lights. Like my the white field scene, I’ve added some color to the light sources to make the scene more interesting. In this one, most of the light sources are blue, since this is a near monochrome scene. Because there is so much blue in the scene, I made the light for the teapot complimentary, so it neutralizes the color of the teapot to read more red. Without that, the red teapot would render closer to purple.

The edge reflection (refl) light could almost be a white card. It mostly just needs enough illumination that it reads white in a reflection, which is different than actually casting diffuse light into the scene. The key light (key_soft) has to be a lot brighter to keep up with the edge reflection values.

The background materials are quite dark, since they are supposed to read as (almost) black, even under the heavy light of the key and bg lights. The diffuse values are in the neighborhood of 0.01 for diffuse contribution and 0.02 for the diffuse color. Essentially, they reflect very little diffuse light. The “table” material has a high amount of reflectivity, which is why it picks up the reflections of the glass bottles.

I adjusted the glass bottle material as well. This material has almost no diffuse and is almost totally transparent. I made these changes because I wanted the bottles to be all reflections, and not pick up any diffuse lighting. I think this looks better on the dark background.

Here are the lights, one at a time:

Background Light (set to diffuse only)

The edge reflection light (set to specular only)

The key light

The teapot “special” light – notice that it’s a little warm to help neutralize the other blue lights in the scene.

Here are all the lights except for the key:

This feels good, but lacks a little interest.

All the lights together!

I recently purchased the 4th edition of Light Science and Magic: An Introduction to Photographic Lighting. It is one of the best books on lighting and surfacing I’ve read. The aim of the book is at photographers, but for people with an intermediate rendering background, it offers a lot of information on how to light (and therefore create) the basic kinds of surfaces we see in the world. Throughout the book there are exercises in lighting, most of which can be done with a very simple studio set up. However, I thought it would be fun to take some of the lessons into the 3d rendering space in order to deepen my knowledge of surfacing and lighting.

The first exercise I took on was that of lighting clear glass bottles on a white “field” or background. This presents a bit of a paradoxical problem. Most people understand that transparent, reflective materials look like what is behind them and what is around them. Lighting clear glass on a white background might seem like an impossibility, since in the white background there is very little to reflect or refract.

The secret to this exercise is knowing that the background outside of the camera’s view must be dark. This requires delicately balancing angle of view and background size -and a lot of test renders. Here’s a shot of the scene set up:

There are three glass bottles sitting on a white, reflective surface. The background is a matte while. There are three lights in the scene. A softbox key light about 45 degrees to camera right, a “sky” fill light above the camera, angled down to the scene and a background light. As you can see, the “table” surface and the background are not much longer than the total scene. My first inclination when creating this image was to use a wide angle lens. I knew that the bottles needed to reflect the blackness of the environment background, and that the wider the view, the more angles I’d see off the bottle. The problem I ran into was with a wide angle view, I needed to have a very long background plane to cover the whole camera view. The background plane was so large, it obscured the dark environment that I wanted to use to define the edges of the glass. I slowly brought in the angle of view and the background plane, until I got just the right amount of black environment reflected. At this point, I was only working with the background light, since I was only interested in illuminating the surface behind the glass, so I could tune the reflections of the black environment.

This rendering is with the background light only. It’s not actually a bad rendering, but it’s more like clear glass on a dark or mixed field. It’s worth discussing the glass material at this point. I started with the Arch&Design Physical Glass. However, I had to tune it a little bit to get the results I wanted. Here’s the modified material:

The main differences between the stock material and my material on this panel are my addition of a little diffuse and a little less reflectivity. This simulates the effect of a little dust on the surface, and helps the glass pick up a little bit more light from the light sources. I’ve set the reflection Max Distance to the extent of my background so there’s no wasted time tracing rays out to infinity.

This panel shows a few key things. The first is the Max Distance for refraction. The smaller this number, the faster glass goes to it’s override color (or blackness). This is an excellent way of simulating partial transparency, since it’s a distance based effect, rather than just a single scaling value. I’ve made the distance a little shorter than the stock 30″, just for fun. I also unchecked the “Color at Max Distance”, since this will tint the areas of the glass that we want to be reflecting the black environment. I’ve disabled the optimization of “Visible area lights cause no reflections”, because I’m using a large key light softbox to create a reflection in the front of the glass. Another touch I added, which makes a lot of difference in any reflective/refractive material is the addition of small surface perturbances. In this case, I used a simple noise with a very small bump amount.

The next light I added was the area key light, which creates a large, highlight on the front of the glass. This evokes daylight coming in through a window, and to the trained eye, a photographer’s studio. As suggested in Light Science & Magic, it might be a good idea to put a loosely spaced grid in front of the light to simulate the mullions of a window. It just depends on the look you are trying to achieve.

The key light also doesn’t make a bad picture all by itself, but it lacks contrast without a brighter background. I put in a third light just for fun. It adds a little color and fill light to the whole scene. I tinted it blue, since people tend to associate blue with glass and water. This is because if we’re outside on a sunny day, reflective objects will reflect the blue sky. In the case of this image, it’s pretty subtle.

You can see the reflection of the large area light square in the angled parts of the bottle. The reflection of a large light is good for showing off the contours of curved objects. As a teaser, here’s the scene with a diffuse material on the bottles:

My normal process for lighting a scene is to put a white, diffuse texture on everything. However, when the scene has a lot of reflective or refractive surfaces in it, this is not a sufficient process. Starting with the reflections and refractions is a better way, since they are dependent on so many other factors. Diffuse lighting, by it’s nature is much less angle and environment dependent . It would be hard to predict that the following scene would look like this, when the transparent objects were added:

Teapot shown for scale.

In 2006 I took a job with Newlands  & Company (NC3D), a visualization firm that does a fair amount of trade in transportation projects.  During my two years there, most of my time was spent leading animation production for one of our larger clients, The California High Speed Rail Association. Everything in this article is from my experience at that time,  on that project. As with any tech company, I imagine things have changed quite a bit in the years since I left.

At the time I took the job, it had been a number of years since I had worked with pre-rendered images. I had been primarily creating real time lighting in games up to this point. The transition was somewhat tough, but strangely enough, one of the primary considerations of both real-time games and animation sequences are frame rendering times. I forget who to attribute this to, but I heard (probably on an FXGuide podcast) that animation studios all have a particular frame time that they are used to, and even as technology advances, they keep pretty near the same frame times. I suspect this has something to do with the expectation a person has for how long an image should take to render. It also probably has to do with how people are used to estimating time on projects.

We primarily worked in 3ds max using Vray as our rendering engine. We used Combustion and After Effects for compositing, as well as a number of support tools like Onyx (for trees), VirtualDub for previewing animation, Autocad for GIS work, and Photoshop for still images and textures. In addition, we had a number of in-house tools created by our technical director.

The lighting for most of the shots was to be idyllic, sunny Californian days. This was a fairly easy lighting task from a creative standpoint. I think the real story of lighting on this project was our movement  from direct-only lighting, to a global illumination rendering pipeline  -something I was not there to see the end of.

When I started, a common outdoor day lighting rig used 3 lights. There was a single, warm direct light for the sun and two, cooler direct lights, usually placed 120 degrees from each other, for sky fill. I worked with this rig for a while, but changed when I learned a great technique after watching Environment Lighting for Production by Tim Jones (Gnoman Workshop).

This image shows our 3 light rig in action. A link to the animation.

The technique he showed was how to use arrays of spotlights for sky dome lighting and ground bounce lighting. The arrays are created by placing spotlights in a dome shape. The lights in the sky array have a cool color and cast shadows down towards the center of the scene. The ground bounce light is created by making a dome array of warm, non-shadow-casting lights below the scene, pointing to scene center. I adjusted the ground dome light positions so that the light would come more from the bottom of the scene than the horizon and set the ground lights to fall off over a specific distance.

I found that by optimizing the number of lights in my arrays, and which of those lights cast shadows,  I could get better render times than using a Vray dome light. The dome light (or environment light) can be a simpler way to work, but I needed the flexibility of the light array to optimize render times. It’s worth noting that noise in image sampling, global illumination, soft shadows and glossy reflections were major hurdles in keeping render times down. For that reason alone, we avoided using GI for most of our animations.

This image shows our dome light rig with area shadows. A link to the animation.

As I learned more about rendering and compositing, I wanted to introduce ambient occlusion into our render pipeline. Ambient occlusion is essentially a monochromatic shortcut for rendering soft shadows cast by large light sources, such as the sky, or the bounce light shadows in a diffusely lit interior. At first, I tried Vray Dirt to simulate this ambient occlusion effect, but I was not pleased the effect. It didn’t render like a sky shadow and it was slow to render smoothly. Mental Ray seemed to have a superior look and faster renders, but there were a number of problems with it as well.

The first problem with rendering an ambient occlusion pass, is that it takes extra render time and set up. This isn’t really a problem if there’s only one render layer, but we had many layers. The second problem we had using Mental Ray to render ambient occlusion was that the Vray Proxies we relyed on for our trees and other high detail objects, didn’t render accurately in Mental Ray. Only the low detail preview geometry of the Vray Proxy rendered in Mental Ray, so the renderings from Mental Ray did not composite accurately with  our Vray renders.

After experimenting a bit, I was able to get the subtle sky shadows of an ambient occlusion render by setting the sky array lights to cast (Vray) area shadows. This definitely added to the render time, but it had a couple of key features:

1)    No need to set up, render and composite additional passes.

2)    Noise control on an area light is pretty simple, compared to noise control in a global illumination rendered animation.

At some point during my first year at NC3D, the growing complexity of our rendering passes made setting up those passes a significant part of our day. The system we had was essentially: The person in charge of the project would remember all the layer sets, Xrefs, mattes, special materials, frame ranges, etc. the animation needed, apply them and kick those off those renders individually. Often times the same xref file would be saved a number of times with a number of different parameters, to avoid duplicating scene data. It could sometimes take an hour or more to set up the render layers, and a single mistake would cause many lost hours of rendering, and artist time. This could be a huge problem, since we often had short deadlines for the animations, and were running near daily renders on multiple projects.

To compound the complexity of kicking off renders, we often had a large number of render layers to manage. Some layers were to help break up the scene to save memory, since we were working on 32-bit systems, with huge landscape textures. Other layers were to provide flexibility in the timing of elements. We had a number of high detail trains in our renders that we needed to time precisely, so it was often better to render these as separate layers. We also had large quantities of high detail cars, and people in the scenes, which needed their own layers as well.  It was not uncommon to have 5 or more different geometry render passes in a single frame.

We looked at off the shelf solutions for render pass management, but our heavy reliance on Xref Scenes made most of those products unusable. The studio owner placed a high value on tools that were tailor made to our process, so it wasn’t difficult to convince him to invest in creating these tools. The technical director spent about two months in development of this tool before we could use it, but once it worked, we wondered how we ever got along without it. The tool continued to be refined while I was there, and became a critical part of our production process.

This is an important lesson for any company that produces art:

Great tools take a lot of resources to develop, but are crucial in empowering artists to make great art. In every project I’ve worked on, the availability, or lack of great tools has made a significant impact on the quality and speed of art production.

Toward the end of my time at NC3D we hired a studio art director that had extensive experience with global illumination rendering for animation. In addition, new features in Vray made GI rendering less noisy and faster to render.  Those changes caused our render pipeline to move almost exclusively to a global illumination model.

This is one of our early attempts at using global illumination. In the animation, the moving objects were rendered with one type of GI, and the static objects with another. A link to the animation.

There were a number of production processes and techniques I wished I had known more about during that job, most of them relating to rendering or compositing. One of the biggest holes in our knowledge was in linear light rendering and compositing. I think we experienced a great deal of frustration in our compositing process due to not understanding linear light, image gamma, and how Vray was handling out of range lighting values.

Part of this problem was aggravated by our choice of the TGA output format, which doesn’t really support color profiles or high-dynamic range values. We tried using EXR files and HDR compositing a little, but we didn’t have the tools to quickly compile dailies and storage space for the EXR files looked like it would become an issue, so the process was abandoned. Had we understood how to properly set up HDR rendering and compositing with liner light, I think we would have seen enough of a visual difference to make it worth integrating into our pipeline.

Related to linear light compositing was our lack of standardization on physical light values and physically-based render effects such as the Vray sky. We simply didn’t have enough knowledge of how physically-based lighting worked and how to integrate it with our varying types of light sources. Currently, there are a number of excellent articles, and books available about liner light and physically-based rendering. It helps that a number of big rendering engines are enabling linear light as the default of how they work. One of my favorite articles on linear light is an FXGuide podcast with Master Zap from 2009.

One of the most important things I learned about lighting at NC3D is that all reflections are the same process of light bouncing. We break lighting into the categories of diffuse direct, specular, mirror, diffuse secondary, etc in order to optimize the rendering process. When I taught myself 3d in college, I understood diffuse direct, specular highlights and mirror reflections as three separate effects. When I was introduced to glossy reflections and global illumination in Vray, I finally understood that all reflections are “mirror” reflections, how they look simply depends on the (micro) surface of the object. As a by-product of this, I realized that a specular highlight is just an approximation of the reflection of a light source. I believe that my misunderstanding of surface rendering stems from the way I learned about 3d, which was: technique before theory. Had I learned more about optics and light when I was learning about 3d graphics, I would have understood what the tools were trying to simulate and known how to use them properly.

For the second installment of my lighting postmortems, I want to start with a brief story from Dance Me Outside, a W.P. Kinsella book – I’ll be paraphrasing, so hopefully he’ll forgive me.

There was a young man with no job skills to speak of. He endeavored to work, so he showed up at a warehouse and told the manager, “I’m the best forklift driver you’ll ever meet”. The manager gave him a working interview, where the young man proceeded to wreck the forklift. Well, he ran away and found another warehouse, where the same thing happened. After repeating this cycle a number of times, he actually became a very good forklift operator.

I told you that story, so I could tell you this one, and how the words, “Let me art direct this game.” Came out of my mouth…

I left Sierra Online after 2 and half years, and got a job at Zombie in the beginning of 2000. During the first few months, I was instrumental in building a few demos that won us project contracts. I was riding high on these accomplishments, so when I found out we’d gotten funding to make a game called Shrapnel. I asked my boss if I could be the art director.

At this point, please refer back to my article on Gabriel Knight 3, where I list the biggest lighting challenges we had on the project. Number one with a bullet:

1)    My lack of experience and hubris.

Shrapnel was a first person shooter game, set in the future. There wasn’t much to it beyond that. Join a network game, play some deathmatch. We were given 9 months to deliver the final product. It’s possible another more experienced team, with an engine that worked better could have made this project. Our crew of 5 relatively inexperienced artists and about the same number of programmers could not make this happen. About 7 months in, the publisher went under and the project was cancelled. I could dwell on everything that went wrong, but really it was doomed from the start. Everyone did the best they could. Sometimes that’s not enough.

I’d like to focus instead, on what I think we did right.

The engine we were using was Lithtech and its associated level editing tool, DEdit. Although we had a number of serious problems with the technology, there were a few excellent features, which we used heavily:

1)    Instanced pre-fab pieces

2)    Texture lights

Building a complete game in 9 months is an incredible challenge. When we started the project, I knew we were going to have to be creative about reusing work, in order to ship on time. One of the ideas I had was to take our 5-6 environment types, and make sets of pre-fab pieces that would match seamlessly together. Lithtech’s pre-fab system was indispensable in this process, since the pre-fabs would refer back to a separate file, instead of becoming part of the level. This allowed us to all work on each of the pre-fab files and main levels, without stepping on each other’s toes. It also allowed me to create lighting prefabs, so that other artists could use pre-approved lighting rigs.

The lighting in the Lithtech engine (at the time) was divided up into lights that effected the static environment and lights that effected dynamic objects, such as players. Many engines separate lighting into static and dynamic because dynamic lights are way more expensive to render than static lighting on fixed geometry. Modern game engines have very complex controls for how each light can affect anything in the scene. The complexity comes from the need to optimize real time lighting calculations as much as possible, so only the features of each light that are absolutely needed, are calculated.

You can see in this screenshot, we’ve got 3 light objects near the light source. One is a spot light source that cast shadows. One source is a point light source that didn’t cast shadows, to simulate bounce lighting. The last light is specifically for lighting the characters.

Texture lights are a brilliant invention, and I haven’t seen them in any other engine I’ve worked with. The basic idea is that there’s an editor that allows you to define lights that are attached to a particular texture. Wherever that texture is placed, the engine knows to render the lights specified in the texture light definition.

This screenshot is lit entirely with “texture lights”.

Another thing I think we did right was to have a single main texture artist for the environments. Jamie Burton painted a large majority of the textures for this project. Not only was his vision for the textures very cool, all the levels had a sense of artistic coherence from his work. As you’ll notice from some of the screenshots, the wall texture detail is pretty subdued. This is important in a first person shooter, where it’s important to be able to pick out your target from the environment. In the scenes that work the best, we put most of the detail into the floor and ceiling.

The skills I was missing that would have had the biggest impact on the lighting quality of Shrapnel are: Lighting Hierarchy and Lighting Color Palette.

Lighting Hierarchy is an important concept that I am just beginning to wrap my head around. The general idea is similar to that of Key, Fill and Special lights that create layers of light and help the player understand the meaning of the scene. For example, the scene has a set of lights that make up the primary illumination for the primary purpose. In this case, it’s seeing your target. The secondary lights then highlight elements like doors, paths and other game-play elements. The tertiary lights might serve to highlight architectural details, signage, or other smaller elements. There’s an excellent article on lighting hierarchy in houses of worship written by Robert Shook and Michael White of Schuler Shook on ChurchExecutive.com.

The lighting hierarchy in this image is a bit messed up. The purple lights in the ceiling should be creating a majority of the illumination. The yellow light strips should be more subdued and act as accents to help the player understand the space better.

Lighting color palette is important to help define the lighting hierarchy, establish mood and give cues to players where there is something that requires special attention. Color theory has never been my strong suit, but there’s a book I re-read occasionally and always get something good out of called: Color Choices: Making Sense out of Color Theory, by Steven Quiller.

He explains all about different kinds of palettes and how to pick colors that will be harmonious or contrasting, while maintaining visual coherency. I highly recommend it.

Mark Long, the co-founder of Zombie, has a history of championing game projects that have a unique visual style, so between Jamie’s textures, the environment artists pre-fabs and some of my lighting, we made a number of really interesting spaces.

Gabriel Knight 3 is a real-time 3d adventure game, developed by Sierra Online, and it was my first job in the games industry, back in 1997. When I was hired at Sierra, I was a year out of college, and most of my CG experience was creating pre-rendered artwork for personal projects. I had never worked on a game and I knew very little about real-time 3d. Lighting has always been a passion of mine, but at the time I had no real experience lighting spaces for games, and not much of a grasp on color theory.

An example of my rendering work, prior to joining Sierra Online. Created in Strata StudioPro.

The art development for GK3 was done using 3ds Max 1.2, and rendered by a custom rendering engine designed by Jim Napier and Peter Freese. We exported our scenes directly from Max, and rendered lightmaps in a custom tool. At the beginning of the project, the game engine rendered our lighting as a multiply pass on top of the diffuse textures. A process I had no experience with, coming from a pre-rendered graphics background. I spent a long time trying to figure out why I could never make my scenes as bright as I wanted.

When Peter came on the project, he introduced me to the concept of a MOD2X render process, which allows for both brightening and darkening of the diffuse textures using the lightmap textures.  MOD is short for modulate, which is the same thing as multiply. When multiplying low dynamic range (8 bits per channel) images, you can only darken, since the values in your inputs can never be above 1.0. The MOD2X helps us get around the darken-only quality of a multiply by increasing the lighting amount in the calculation by 2. Here are the equations:

MUL Lightmapping: diffuse*lightmap = final surface

For example, if the diffuse texture brightness is 0.5  and the lightmap is 1.0 in the light areas and 0.25 in the dark areas, the final surface will range between  0.5 and 0.125 brightness. In the light areas, the final surface doesn’t change value at all and in the dark areas, it gets darker.

MOD2X Lightmapping: diffuse*lightmap*2= final surface

With the same diffuse and lighting values as the MUL example, the range becomes 1 to 0.25, so it’s doubling the lighting values compared to MUL only.

The downside of MOD2X, is that it becomes possible to create a final surface that is brighter than the rendering engine can handle and will be clipped to 1.0 (white) during rendering. The lighting artist always has to balance texture albedo with lighting values to make sure lighting in the scene doesn’t clip to white in an unnatural way.

After I knew about MOD2X, I wondered why anyone would use MUL only lighting. I think part of reasoning behind using MUL was a graphics hardware issue, and the other part it was a judgement call on how the textures should be authored. With MUL only lighting, you make your textures a bright as you ever want them to be when fully lit. With MOD2X you would author your textures so they could be both lit and darkened without going to pure white or black.

For every project, the lighting must have a primary goal. On GK3, atmosphere was the primary goal. We were building a slow-paced mystery-adventure game that was supposed to be spooky, so we needed the lighting to create that feeling. It was less important to have easy target recognition, or a sense of high speed, or some other game-play related goal.

Because of the 3^rd person view of the game, it was important that the walls, floor and to some extent, the ceiling to be lit with equal care. Not every game requires the same amount of attention to each angle. For example, in a first person game, the walls will be the most important surface to light, since they will be the largest part of our view.

The lighting rigs I worked with on GK3 were designed by our art director, Ron Spears. For outdoor scenes we had two direct lights. One for the sun, and one for the sky. The sky light was placed opposite the sun angle and perhaps a little steeper depending on the scene. For artificial lights, we’d use point or spot lights. These were all standard light sources in Max.

The main lighting challenges we had during GK3 were:

1)    My lack of experience and hubris.

2)    Inconsistent/no art direction because of turn over.

3)    Limited lightmap resolution due to video card requirements.

4)    Creating lighting consistency between a large number of scenes.

Though I knew very little about lighting environments when I started the project, these are the areas of knowledge that would have had the most positive impact, had I known them:

Using lighting to assist with way-finding. This means using light to help the player know where to go and what to look at. Often times, the artist must put in special lights that are cheated from the key light direction in order to highlight an important surface or object.

Lighting color temperature. We needed more consistent and better thought out values for all our lights. Many artists tend to think of light colors as absolutes, but our brains are constantly trying to neutralize the colors we see. A big shadow that seems blue when we look at it, will desaturate when we’re standing inside of it. I don’t know of many rendering engines that will do this kind of compensation. Because of this, lighting colors are either too saturated or not saturated enough. It can help to define light color ranges that are different depending on the color of the key light in the scene. Warm lights that are used with cool daylight will be more red and saturated than when warm lights are used exclusively. We also missed a chance to influence mood further by using the light colors thoughtfully.

Creating strong directional lighting to bring visual depth to a scene. This is not always necessary depending on the lighting situation, but it can increase how easy it is for the player to “read” the space. I feel a lot of our scenes were visually flat, and really needed more play between light and shadows. In animation and stills this is easier, since you have a key light that will be used for that particular scene. In real-time 3d worlds, the player is moving around all the time, so a key direction that works from one set of angles will be sub-optimal for other areas in the scene. One way to make this easier is to design the scene geometry with this in mind, and allow opportunities for directional light indoors.

This seems like a no-brainer, but anything lit by the night sky will be darker than the sky. All our night skies were too dark.

My own misplaced confidence in my abilities might have been my worst mistake of all. The person I worked most closely with, Chris Brockett was very good and lighting and texture work, but I was so wrapped up in my own vision, I missed an excellent chance to learn from him.

Although this is one of my more successful lighting set ups from the game, you can see a number of the flaws: The lighting does not clearly indicate what’s important to the player, the sky is darker than the tree silhouetted against it, the street lamps cast distracting shadows, and the colors are not very evocative.

Chris Brockett, the other main environment artist on the project had a far superior grasp of light and color than I, which you can see demonstrated in this screenshot of the temple.

We had a few different art directors during the course of the project. When I came on Ron Spears was the art director. Later on in the project it was Mark Peasley, and then Richard Hescox came on at the end. Keeping consistency in the art was difficult with these changes. However I learned a number of good things about lighting from each art director. Most notably:

1)    Open all the textures for your scene at once, and make sure they all are in the same palette, including value and saturation. That will help the scene will light consistently.

2)    When lighting multiple times of day, pick a different palette for morning and evening. In GK3 we pushed our morning sun to green and our evening sun to pink.

Limited lightmap resolution is not always a killer when it comes to lighting. Interior spaces don’t often need a lot of lighting detail because a lot of interior light casts soft shadows. However, when lighting outdoor scenes in bright, direct daylight, the lack of crisp shadows can be hard to work around.

We did a few things to help the lighting be more consistent throughout the game. I made a single outdoor lighting rig for each time of day, and an orientation map for each scene so that the sun angle was consistent across connected scenes.

For further reading on the environment art development process of Gabriel Knight 3, here’s an interview I did with Nico Sels, who runs the SIDNEY web site, dedicated to GK3:

http://sidney.fwheel.net/t_aubrey.htm

I’ve decided for the new year, I will write lighting post mortems for some of the projects I’ve worked on. I hope that these will help other artists learn from my mistakes and provide some insight on different lighting work flows. I admit, some of the projects are awfully old, but I think it’s also nice to remember how far we’ve come as artists in the digital age.

I recently moved my photo work to the domain www.aubreypullman.com, and moved my 3d work to the root of www.rendermack.com. I’ve always thought of Rendermack as specific to my CGI work, since it’s a play on Renderman + Mack Daddy. It also seems more traditional for photographers to use their own names in a professional context rather than an internet nick, but perhaps that’s a generational thing. However, I’ve made just one public account for tweeting about my professional work (http://twitter.com/rendermack), and most of my Instagram photos get posted there. I figured it would be a nice way to encourage people to visit both sites.