Lighting Tutorial - Francesco Giroldini – April 2010

Hello friends,

This tutorial describes a way to break down a shot lighting in a simple, step by step way.

Model and textures kindly provided by the talented Stevie Lewis http://chocosweete.blogspot.com/

Before you get started make sure you are looking at this pdf the way it was intended to.

Click on “start” > “all programs” > “Adobe” > “Adobe Acrobat 9 Pro”

Once Acrobat opens, click on “edit” > “Preferences”

This will open the preferences window.

Click on the “Color management” button and under the RGB working space drop down menu make sure sRGB is not selected, if it is, then choose the monitor color space (in this case HP LP 3065).

Press ok.

Close Adobe Acrobat.

This is the image we’re starting from:

And this is the image we’re shooting for:

I talked to Stevie about the her intentions for this shot, and she told me she meant it to be cheerful and optimistic; she meant this shot to be happening outdoor sometimes around noon or early in the morning.

I think overall she did a very good job portraying this time of the day, although I would have added a little bit more variety of colors and values to the sky, which right now I believe looks a little flat.

When I start lighting a new shot I usually like to start fresh and hide all the lights but the environmental light, just to take a look at how the shot looks under a neutral lighting condition.

If you’re not familiar with how to create an environmental light here are some tips:

Open render settings and under the “Features” tab, right click on the “Environmental Light” slot and choose “Create Renderman Environmental Light”

By default the environmental light is going to use raytracing in order to perform the occlusion, so make sure to have Ray Tracing enabled.

For most scenes I would recommend these raytracing settings:

Raising the “max ray depth“ and “max specular depth“ tends to raise render time quite a bit and might not be necessary unless you’re working with a scene file that involves many reflective or transparent objects.

Rule of the thumb for these setting is to keep them at 1 and not raise them unless you notice a dramatic improvement in the quality of the final image.

“Max Diffuse Depth” has to do with the number of bounces for ray traced color bleeding; 1 is good.

By default the Environmental light is not going produce good shadowing; in order for it to look nice and clean it’s going to be necessary to change the sampling method from “filtered” to ”sampled”, increase the samples from “64“ to “256“ and the max variation from 10 to either “1” or “2”.

The max distance value is going to depend on the scale of your scene file, I usually recommend something around 50, but if your scene scale is really small you might need something like 10,or if your characters are gigantic then go for something like 200.

Right off the bat I can see that something is wrong with the render; there should be some occlusion between the hair and the skin and more shadowing happening on the shorts.

After a short investigation I realized that visible in reflections and refractions was turned off for the shorts and the hair, I turned it back on and rendered the image again.

You can see how by turning these attributes on we got back some shadowing on her face and on her shorts. The eyes have also visible in reflections and refractions off and that’s a good thing, because raytraced occlusion doesn’t work well with transparency.

If we were to turn visible in reflections and refractions on on the eyes, this would be the result:

Super creepy right?

First step for blocking in the lighting is setting the mood with the least amount of lights possible; outdoors shots are usually easier than indoor ones because the environmental light does a good job simulating a sky dome around the character and providing a nice soft fill light.

A personal preference of mine is to change the shadowing method of the environmental light from “occlusion” to “color bleeding”.

The reason that makes me prefer Color Bleeding to Occlusion is that Occlusion tends to add a layer of gray haze into the shadows, as if you were to try and darken a region of a painting by mixing black into the colors. The result is that usually occlusion tends to kill the organic look of an image, while color bleeding tends to amplify the feeling that all these objects are sitting next to each other.

Left: Occlusion Right: Color Bleeding

In order to change the shadowing method to “Color Bleeding” select the environmental light and click on the Shadowing drop Down menu and choose ‘Colorbleeding”

By default Renderman is going to use raytracing in order to perform color bleeding, and we don’t want that, because it’s so slow, you’re going to grow a beard by the time you render is finished.

So what we will do is switch to a point-based color bleeding computation:

Right click under the bake shadowing slot and choose “create render radiosity”

Makes sense?

Probably not right? I think this step requires some extra explanation, just to make sure you understand what happens under Renderman’s hood, and why this speeds up render time so much (compared to the raytraced method of computing Color Bleeding.

What is a rmanRenderRadiosityPass ?

When you right click on the bake Shadowing slot and choose “create render radiosity”, Renderman creates a new pass, which is going to be computed BEFORE the image is rendered.

This pass consists in taking every geometry that is visible on camera and splitting it into small little dots that share the same shading information as the original models; these clouds of dots are conveniently called “Point Clouds”.

You can imagine the point cloud as a volumetric cloud of light that brightens whatever gets into contact with it.

Above are some snapshots I took of the point cloud for this shot, ( usually located inside the “Renderman/data” folder) . In order to look at the point cloud file, you need to open it using the ptviewer.exe application that is located in:

C:\Program Files\Pixar\RenderManProServer-14.3\bin\ptviewer.exe

After this pass is computed, the final rendering pass starts and at this stage all these little dots become small light sources that light the surrounding geometry inside the distance set under the max dist. attribute.

The brightness of their light is going to depend on the shading information that was stored in the radiosity pass.

Enough for the theory, let’s keep lighting this shot!

Something that generally happens when switching the shadowing method from occlusion to color bleeding is that the image gets overall brighter. This is because now the local color of the objects is taken into consideration in the shadowing computation, and usually that causes bright objects to look a bit too floaty.

For example, if we look at Stevie’s face it becomes evident how much shape was lost. In order to gain back some of the shape we can add something called “Environment Controls” on the shader.

Select a material and open the attribute editor; now click on the ”attributes” tab and under the “Renderman” tab click on “Add Environment Controls”.

This will add 4 new attributes at the bottom of the material, under the “Extra Renderman Attributes” tab. If you scroll down to the bottom and lower the radiosity scale to 0.5 this will cause the face to emit less color bleeding.

Occlusion Color Bleeding Color Bleeding/radiosity scale 0.5

The difference might not be huge, but by altering these controls on the face material we managed to bring back some shaping on the eyes and teeth.

Let’s go back to the material attributes and let’s lower the “Color Bleeding Scale” to 0.5

Color Bleeding Color Bleeding/Radiosity scale 0.5 Color Bleeding/Radiosity scale 0.5

Color Bleeding/ColorBleeding scale 0.5

Again, the difference is not huge but we now managed to get to a solution that gives us both the benefits of Occlusion and Color Bleeding combined.

Alright! Enough for the fill Light! Let’s start working on our key light!

The key light, as the name suggests is the single most important light for a shot; it’s usually associated with being the brightest light source illuminating the set or the character.

The key light main purpose is to give shape to the character/set and bring out the features of a model.

I added a spotlight and positioned it about 30° to the left of the character hoping that it would bring out Stevie’s features; the result is less than ideal.

All of a sudden all the subtle tonal variation and details in the shadows has been erased and what we’re left with is a flat looking image. There are a couple of reasons why this happened, the main reason being that the keylight is not casting any shadow, and therefore is brightening the image where it shouldn’t.

I’m gonna have the light cast a depth map shadow.

Keylight, no shadow keylight, depth map, resolution 512

This already looks a little bit better, although a couple of things still bother me about the image to the right: Her left side of the face looks a bit too flat and a distracting shadow from her hair appeared over her eyebrows. Also her whole body is in shadow, and I can already tell that we’re going to see a lot of distracting shadows as she starts moving.

Luckily enough the spikes are a separate geometry so I can easily disable cast shadow on them.

This is slightly better, but the distracting shadows on her body remain.

What we can do is select all of Stevie’s body except the head, hair and teeth, shift select the light and break the light links (under “Lighting/Shading > Break Light Links), then duplicate the key light, disable “Illuminate by default” (this will effectively turn the light off and force it to light only what is specifically light-linked to it) and then link this second light only to the body(under “Lighting/Shading > Make Light Links).

Now we will lower the shadow resolution on this second light, from 512 to 128 so that the shadows on her body are going look a bit softer than the one on her face.

A very important step for the shadow to look at this low resolution is to convert it from “Depth map” to “Deep Shadow”:

Select the light and in the attribute editor click “Attributes” > “Renderman” > “Add Shadow Attributes”

Although this trick will force us to compute 2 shadows, it will give us more control over the look of the final image.

I will now add SSS to Stevie’s model hoping it’s going to help adding some reds into the shadows.

Select the face material and under the “Attributes” tab click on “Renderman/ Add Subsurface Scattering”

This will add a couple of extra attributes at the bottom of the material:

SSS can be expensive, so it’s good to be able to optimize it as much as possible.

Right click on the “Scattering Map” slot and click on “Create SSOrganize”

This will create an SSS pass that can be used to set the SSS to either compute or reuse.

This is particularly important when doing test renders, because it allows you to reuse the SSS instead of recomputing it every time. Keep in mind that the SSS is computed per frame so setting it to reuse will only save render time if you keep working on the same frame.

Here’s the render with SSS.

While before the scene was rendering under 2 minutes, now render time jumped up to 8 minutes.

This is unacceptable but luckily enough there’s a simple way to gain back some render time.

Go back to the SSOrganize pass, and then click on the tab called rmanSSrenderPass1.

Right click on the object set slot and choose “create object set”; this will create a new object set that is going to be accessible from the relationship editor. The set is going to be empty by default.

To access the set editor, click on Window/Relationship Editor/Sets

As you can see the newly created set is going to appear in the list; select the set.

Now we’re going to add the geometry that we want to receive SSS in the set, by finding them on the right column and clicking on them. I’m including Stevie’s face and hands.

Once you’re done with your selection just close the window.

The images should look pretty much the same, but now it should render under 5 mins

Although we still have a few lights to add, I think it’s time to look at the render in context; in the end I’ll probably take Stevie and render it separately from the background just so I’ll be able to integrated her better, but for now I’ll just place the image of the background behind her and render it in Maya:

The result is less than exciting; she doesn’t fit the picture at all; her black are way too dark and there’s no sign of bounce light coming from the ground on her face.

To brighten her up a little bit we could of course try to add a couple of fill lights and such, but that would probably end up washing her up too much. A simpler solution involves tweaking the gamma of the image, using something referred to as “linear workflow”.

Go under the “Passes” tab, in render settings and change the second slot under “Exposure”

From 1 to 1.6: render the image.

As you can see the resulting image looks now overall brighter, but a bit too washed out.

To bring back the contrast we are going to have to gamma correct all the color textures in the scene. This is potentially a time-consuming step, but it only needs to be performed once, so it’s not too bad.

The easiest way to do this, is to use this script: gammaTools_1.0

Instructions on how to install the script are included in the zip file.

To execute the script just select a texture and press the button on the shelf:

When executing the script a gammaCorrect node will be generated and connected between the color texture and the material. The default gamma corrections is going to be .455, you will need to change that to .625 on all channels. The reason for these odd numbers is that they depend on the gamma correction applied to the whole picture, and since we applied a gamma correction of 1.6 (as opposed to the standard 2.2) we are going to change the default value of .455 to .625.

Practically, what we are doing is washing out all the image using the 1.6 gamma correction and getting the contrast on the textures back using a counter-color-correction of .625 (1/1.6= 0.625).

If by any chance you run into some problems installing the script, you can easily do this by hand by creating a gamma correction node, changing the values from 0 to .625 and plugging that between each color texture and the material.

After applying the gamma correction nodes to all the textures in the scene file, render the file.

If we compare the image before the color correction and after, we can see how now the character looks a bit brighter now and integrates a bit better with the background, all of this without adding any light or increasing render time.

We are now going to add a bounce light to the scene, just to get the point across that it’s noon and light is hitting the ground and bouncing up. Since the ground is surrounding Stevie, I’ll create two spotlights

One for each side, and since I don’t want each side of her face to be equally lit, I’ll make the right bounce light slightly brighter and of a slightly warmer hue.

When dealing with fill and bounce lights, it’s important to make sure that those lights don’t wash out the image and ruin the shaping of the key light. Something I see people do a lot is creating ambient lights or directional light that cast no shadow, just to brighten up the whole image. Although these methods will brighten up the image, they will also tend to kill the contrast, by raising all the blacks in the scene and washing out all the contact shadows.

Bounce lights are traditionally very soft light sources, so they’re generally achieved by using dim lights with very soft raytraced shadows. In this case I achieved the same look by using a deep shadow with a very low resolution instead.

Below are the settings I used for the bounce lights.

I believe we did enough work inside Maya, it’s time to bring this image into Nuke and start playing with our options; later on I might decide that I still need to do some work inside Maya, but for now, I want to get a feel for the final look of the image.

In order to composite the background image behind Stevie, we are going to need to have an alpha.

The environmental light by default is going to be visible in the renders, so in order for us to be able to save an image of Stevie with an alpha we are going to have to check “primary visibility” off on the environmental light. This won’t make any difference in the illumination.

Alright, let’s open nuke and let’s composite Stevie over the background.

I’ll go over the basics of Nuke, for those who haven’t played with it before.

To import a file press “r” inside the Node Graph window.

Navigate in the window and grab the background image.

The image will appear in the node graph window.

Repeat for the foreground element.

To navigate the node graph window, press alt and middle mouse drag to zoom in and out, alt and left click to pan around. Press “f” to frame the view around the selection.

The single most important setting we have to change before we start working is the “full size format” setting.

This is the size at which all the elements that we are going to create inside Nuke will have and it can be accessed by pressing “s” in the node graph window. If we leave this at default (640x480) elements such as beziers and constants may not match the size of the input images and strange artifacts such as streaks will occur.

I see a lot of people freaking out when running into this issue, so I always recommend double checking the project settings immediately after importing the first few files into the node graph window, and making sure that the size of the input images (in this case 853x480) matches the size set in the “full size format” slot.

The frame range slots are also very important to keep an eye on; for example if you were to import an image sequence that starts at frame 120, you would need to adjust the project settings to start at 120 as opposed to 1.

Alright, now that we are all set, let’s have some fun compositing this shot.

In the node graph window, select the foreground image, shift select the background image and press “m”. This will create a “merge” node, to see the resulting image, click on the merge node and press “1” in the node graph window.

The resulting image looks funny, Stevie is out of place! What happened!

This is the typical result of merging 2 images that have different sizes, as you can see at the bottom right corner of the picture the background resolution is 2000x1125, while the resolution at which we rendered Stevie is 853x480.

The simplest way to fix this is by adding a “reformat” node between the background and the merge node.

Select the background image; press the “tab” button on your keyboard (the one above the caps lock) and type “reformat” in the little box that should appear. This is a simple way to add nodes to your project; just press “tab” and type the name of the node you want.

Since we previously set the project settings correctly, the reformat node will automatically know that we are working at 853x480 and it will resize the background image to fit.

Alright, next step is going to be adding a “lightwrap” node. This is a really cool node that allows the foreground to blend a bit better into the background

Deselect any node, press tab, and type “lightwrap”

As you can see the lightwrap has 2 inputs” A and B; A is for the foreground, B is for the background.

just click and drag on each of the2 arrows and release on the foreground and background.

The node graph should look like this now:

Select the lightwrap node and press “1” in the graph node window; double click on the lightwrap node so that the properties panel will load the lightwrap attributes.

“Enable Generate wrap only”

Increase the “Intensity” to something higher than 0

What you should see is a blurred version of the background spill into the alpha of the foreground.

Now we’re going to do another merge between the lightwrap and the A input of the previous merge.

Deselect any node that might be currently selected and press “m”; this will create an empty merge node. Double click on it.

The merge node allows different methods of overlaying 2 images on top of one another; by default it simply allows to stack an image on top of one another, but by changing the operation to “screen” it’s possible to achieve a different result.

Alright, now that we set the merge node for a “screen” operation we have to connect the lightwrap as input A and the plate of Stevie as input B.

This will add the lightwrap on top of Stevie’s render.

Now we’re gonna have to drag the arrow that goes from Stevie’s render to the “Merge1” node and reconnect that to the “Merge2” node. That means that while before we were merging Stevie over the background, now we’re merging Stevie + Lightwrap and the background.

Select the merge1 node and press “1” in the graph node window.

To check what difference the lightwrap makes, select the “merge2” node and press “d”; this will temporarily disable the node. Press “d” again to enable the node again.

Double click on the lightwrap node again and check the “disable luminance based wrap” checkmark

This is usually a checkmark that I like to keep disabled, but on certain shots, where the lighting is not very dramatic, it can help integrate the foreground a bit better into the background.

Notice how some green now spills into her t-shirt, blending her better with the background.

One last touch that I usually like to give to a picture is some sense of gradient or hint of where the light is coming from; I usually create this by adding a very soft flare on top of the whole image, trying to mimic the look of camera flare.

Another small thing I will add to the shot before calling it done, is a little variation in value on the background.

Select the background plate and press “g”; this will create a grade node.

Grade is good for darkening or brightening an image, tinting it or adjusting its gamma. The very nice thing about the grade node is that it allows a mask to control where the effect has to happen.

Double click the grade node and change its settings to:

The mix slider at the bottom allows you to quickly blend between no effect and full effect of the node.

Deselect any node in the scene and press “p”; this will create a Bezier.

Select the little tail in the shape of a triangle coming out of the grade node and drag it over the Bezier.

This will set the Bezier to the mask for the grade node; the image will revert to what it used to look like before the grade node was applied, because the Bezier node is empty since we haven’t drawn any shape just yet.

Double click on the Bezier, and press ctrl alt and click on the image to start drawing points.

If you can’t see what you’re drawing you might have to press “o” to enable the visual guides.

If you press ctrl and drag over a point you can soften that point; be aware that beziers are by default set to “autokeyframe” so if you move from frame to frame and keep changing the shape of the Bezier you might get unwanted animation.

This is the image after the Bezier has been drawn.

As very last finishing touch I’ll add some little grain.

Go to the very end of the node tree, select “merge3” and press tab; type “grain”; if you can’t see a change, press “1”

The noise is only visible on Stevie now; the reason for this is that by default the noise is only visible on the opaque portion of an image, and because the background images is a jpg and doesn’t have any alpha, the noise is not visible on it. This is ok for me because the background already has some sort of texture to it and probably wouldn’t benefit from noise anyway.

By default the grain is pretty heavy handed; a simple way to adjust its intensity is to create a merge between a grained version and a grain free version and use the mix slider to blend between one and the other.

Press “m” to create a merge node.

Connect the Grained image to the A input of the “merge4” node and the “grain free” image to the input B of the “merge4” node.

Double click on the “merge4” node and change the operation method to “atop”; this is going to ignore the fact that the background has no alpha and its going to comp them as expected.

Adjust the mix slider as needed.

To export the images, just select “merge4” and press the “w”;

Click on the folder icon next to the “file” slot and type the path where you want to save the images, followed by the image name, numbers of digits and extension; for example:

M:/MAYA/projects/second_project/images/Comped/Stevie_Comped.####.tif

The four “#” are the same as the “frame padding” option inside maya; typing four of them means the images will be renamed “image.0001.tif”, “image.0002.tif”, etc

We don’t need to include an alpha channel in our output, so we can keep the alpha set to “none”.

One last thing to double check is the compression type, by default set to deflate, which should be set to LZW instead.

When you are ready just press the render button and render away!

I hope you folks enjoyed this tutorial; for any question, suggestion or complain feel free to e-mail me at

Giroldini@gmail.com