Making of the patient
The following article isn’t intended to be a tutorial as such but rather a focused ‘making of’ during which I will attempt to explain a lot of my working practices and techniques. I’m going to illustrate this with Videos and images where appropriate. I’ve also included at the bottom of the article a link to the scene file, which includes a version of the scene with a frozen mesh and all the textures in JPG format at full resolution. I have no intention of discussing modeling or sculpting techniques, as there are already thousands of head modeling tutorials out there and I really don’t think there is anything I can tell you that you wont find better demonstrated elsewhere. Instead I want to focus on the Texturing, Shading, Lighting, Rendering and grading of this image.
There are various ways to texture a face, a lot of people tend to use 3d paint programs such as Zbrush or Body Paint, some artists still prefer to paint straight onto the unwrapped UV layout, however I prefer to use a Projection mapping / texture-baking technique. The process involves camera projecting a reference photograph straight onto the mesh and then baking the result out to the UV map, this allows me to create a distortion free base from which I can produce my colour, spec and bump maps.
Setting up projection scene
Setting up the scene is a fairly simple process, for this project I used two cameras, one for the front and one for the side. When setting up the cameras I always try to use the longest lens possible, this helps to flatten out the features of the face and reduces texture distortion due to perspective See Images and video Below
Above Left Lightwave camera layout and settings Above Right Flash video detailing the processes involved in setting up the scene
Above: an example of what the front and side renders should look like
Now I want to show you a technique that I use to match features on a photograph to the features on your mesh without resorting to the clone tool. For this particular project I modeled the face using the photographic reference as a guide, therefore the features are approximately in the same position as those on the texture reference. However they are not exactly right so I used the “Liquify” tool in Photoshop to gently nudge the features on the photograph around until they align to those on the mesh. Take a look at the video below to see how it works
The next step in the process is to re-project the adjusted reference material back onto the mesh and bake it into two UV maps front and side which can later be combined to create a base for the colour map. In order to re project the textures I used exactly the same scene and cameras as I did to render out the front and side views. The video below shows the how the process works.
Above: Surface settings for front projection
I Set the surface of the mesh to that shown in the above image using the front facing camera to project the front image. The next step is to use Microwave to bake the front projection onto into UV map. Use the settings show in the image below and make sure you check the “Bake Colour” tab and set the Expand to 3 pixels, not including a 3-pixel border may result in visible seams around the edges of the UV maps. Also something that I’ve noticed with Microwave, if you don’t set the resolution of the scene camera to something equal to or higher than that of the texture your baking out then it will blur the baked image. Once Microwave has done its thing and I’ve saved out the baked front image I then switch to the side camera apply the side projection image to the surface and run through the whole process again, not forgetting to change the projection camera to ‘side’ in the surface settings.
above Left : Microwave settings Above Right: Flash video depicting the reporjection process
If you don’t have the microwave plug in you can use LW 9.2’s new Surface baking camera. Its pretty much the same except it doesn’t give you the option to bake out colour only Apply the textures using the same technique as above however make sure you set the diffuse to 0% and the luminosity too 100% this will give you a pure colour output. By this stage you should have two textures that look like something like those shown below
above Left : Baked front projection Above Right: Baked side projection
Combining baked projection maps
In order to combine the above images into a single map I loaded up the projection scene and positioned a spotlight directly down the x axis and set it to cast ray traced shadows. the purpose of this is to cast light onto only the front facing polygons. I then used Microwave to bake an illumination pass into the UV map. This gave me a black and white image where the front facing polys are white and the side facing polys are in shadow and therefore black. To combine the images I loaded both front and back maps onto individual layers in photoshop the front on top. I then loaded the baked illumination pass onto another layer and used the ‘Select / Colour Range’ tool to make a selection out of the black / shadowed areas of image i.e. all the side facing polygons. This selection was then used to delete the stretched side facing area of the front map. this technique is better illustrated in the video below.
Removing the specular highlights is a fairly simple if not time-consuming process. I did this by creating a new layer in photoshop set to darken. Using the clone tool set to 50% opacity and sample all layers activated. I selected an area close to the highlight that I wanted to remove and cloned over the top. As you can see from the video because the layer is set to darken, only the areas lighter than those being cloned are affected. Generally specular highlights are fairly bright in comparison to the rest of the skin tone so only the small spec areas are cloned over.
Above left: Photoshop layer settings Above Right: Video demonstration of spec removal
Zbrush touch up
For this project I used Zbrush to touch up the texture map, generally this is all I use it for, I do all my displacement painting in Mudbox, I know its slower and cant handle as many polys but I prefer the interface and in my mind it’s a much more user friendly piece of software. Before exporting an obj from Lightwave you should first make sure you have a map applied to face using the UV map as Lightwave doesn’t export obj UV maps unless there is a texture applied to the model.
The spec map is in my opinion the easiest one to get very wrong, I’ve seen a lot of very nice models ruined by plastic looking skin due mainly to a poor spec map. I have a pretty simple way of generating the correct shades for the correct parts of the face. This is demonstrated in the Video below. Basically it involves rendering the model with a single spot light set to 100% and applying a single grey colour node to the spec channel in the node editor. I then do an FPrime render of the face and adjust the grey until I think the spec looks about right on the cheeks, which are probably the least reflective part of the face. This grey then forms the base colour for the map. Next I move onto a more oily area such as the tip of the nose. Focusing only on this area I adjust the grey again until the spec looks right, this colour can then be used to paint the end of the nose in the spec map. I repeat this process for the forehead, chin, lips, ears etc. once again its probably better illustrated with a video.
In most cases I would make the gloss map in exactly the same way as I would the spec using the above process. However this is the first time I had used lightwave’s node based texturing system. So I cheated a little bit by simply plugging the spec map into the gloss channel and putting a Colour Tool node in-between, this allowed me to adjust the brightness and contrast of the spec map. As a rule the more oily and reflective the skin is the tighter the highlights therefore the spec map can be used quite effectively with this technique.
Epidermal / Sub dermal
In order to use the simple skin shader node I had to make two more maps Epidermal and Sub dermal. The epidermal colour defines the skin colour without any blood. I made this by simply de-saturating the colour map. The sub dermal map should approximate the layer of fatty bloody tissue below the skin. I made this layer by adjusting the curves, increasing the saturation then blurring the image, because it only defines overall colour of the skin there is no need to have a lot of detail and things like hair follicles and eyebrows can be completely removed using the clone tool. Its also important to remember that on areas such as the bridge of the nose, chin and forehead there isn’t as much fat under the skin so it might be worth going over the map with the dodge tool and de-saturating these areas slightly.
Above: Maps used with the simple skin shader
The Bump map is one of the easiest textures to make, assuming you have taken care when creating the colour map. Its important to get this one right, I’ve seen so many fantastic models ruined by terrible bump maps. The quickest way to make a bump map is to load in the colour map, de-saturate it, then use the high pass filter as shown in the video, after this use the curves tool to adjust the map until there is a decent amount of contrast between the greys. If you need to add more detail to the wrinkles you can go in as I have on this map and paint them on. I like to do this for the areas under the eyes and the lips. For the eyebrows, which are generally dark you will need to invert them. I found the best way to do this is lasso them cut them into a new layer, invert them and change the overlay to screen.
As with the modeling I don’t feel its necessary to give you an in-depth rundown of my Mudboxing techniques, there are plenty of digital sculpting tutorial videos out there, and this is the first time I’ve ever really used it on a project so I cant really offer you anything that I’m sure a far more experienced digital sculptor could. I have however included a few videos showing the various layers in Mudbox and also the export settings that I used to get the displacement map into lightwave.
Above left: exporting model from Lightwave for Mudbox. Above right: Displacement layers and map export in Mudbox
Displacement node setup
As you can see from the video below I’m using the shader node system in LW to apply the displacement map, I did consider simply exporting the high-density mesh from Mudbox but eventually I want to animate this mesh so I really needed to keep the low res geometry.
I started this project with the intention of trying out Lightwaves node based shader system and more specifically the simple skin shader. I had done some work in the past with MISS shader in Maya and was becoming increasingly frustrated with G2’s flaky performance. The Simple skin shader is very similar to MISS in Maya, its fully supported by FPrime so its really simple and easy to make changes and see them update immediately even with multiple bounce radiosity enabled. Before I start applying textures and shaders to any character model I begin by setting up the scene with a single distant light set to 100% if you can get the skin to look good with a simple lighting setup like this its only going to get better when you add some radiosity, rim and spec lights later on.
Above: Node setup as described below
This channel is used to plug the base colour map into I used a setting of about 50%. Diffuse Epidermis and Sub-dermis Visibility should add up to 100%
This channel controls how much diffuse shading there is on the bump map
The colour of the specular highlight, I always use a pale blue for the spec colour
Plug the Spec map into this channel. As you can see from the diagram I have placed a colour tool into the flow this allows me to adjust the brightness and contrast of the image just in case I need to fine-tune the spec later on. Don’t forget that if your plugging a colour map into a scalar channel like spec then you should use a colour scalar node as show in the flow below.
For the Gloss map once again I’ve used a colour tool node, as I explained before I’ve used the same map for the gloss and spec but cheated but adjusting the brightness and contrast of the gloss until it looks correct, this is made very simple by using FPrime as you can see the updates on the rendered image almost immediately, it also saves a lot of time messing around between Photoshop and lightwave. This type of cheating is just one of the many advantages to using the node shader system. It’s much better
This setting allows specularity to vary according to the viewing angle of the surface, I’ve set it to 35%
The refraction index is a measure for how much the speed of light is reduced inside the medium in this case the light traveling through the skin. The refraction index of skin is generally believed to be between 1.37 and 1.42 this shader had it set to 1.37 when I loaded it up so I just left it where it was as it seemed to work pretty well.
The epidermis is the colorless translucent top most layer of tissue; it allows a lot of light through to the fatty sub-dermis layer below. I’ve set the visibility to 30% the distance to 1.5mm, as is the default on the shader. When changing the distance settings you should always take into account the scale to which you model is made. I modeled my head to a realistic scale and as such the default settings were pretty much right from the offset. Gamma I l left at 1.0 and I changed the samples to 4, simple for speed at first but when I did some render tests I couldn’t really see any difference between 4 and 8 so I left them at 4.
This is the relatively thick fatty bloody layer below the Epidermis it contributes quite heavily to the reddish colour of the skin as the light passes through the capillaries where the light is tinted by the blood. I’ve set this to 20% make the overall value of the Diffuse Epidermis and Subdermis up to 100%. Again I’ve left the Distance at the default 2.2mm and the gamma at 1.0 and as before changed the samples to 4.
The intensity of the bump map, I’ve increased it too 150 percent, but this value depends on how much contrast you have in your bump map the only real way to figure this one out it to do some renders and see what looks good.
Advanced / Reflection settings
By default the simple skin shader uses reflection as well as specularity, however I really didn’t want to use reflection for this image. You are more than welcome to fiddle around with it and try to make it look good, but I kept getting rendering artifacts in FPrime and it was taking far too long to render out the reflection noise with multiple bounce radiosity turned on so in the end I just went for spec. If like me you decide to turn the reflections off you have to set the Mode to Spherical Map and reflection image to (none)
Lighting / Rendering / Camera Setup
This is my favourite part of the whole process it’s where all the laborious modeling, texturing finally pays off. Personally I could spend days fiddling with lighting rigs and settings. For this image I wanted something that would show the face off but remain fairly gloomy and dark. I wanted him to look lost and alone. In the end I went for quite a stark contrast between light and dark keeping his entire left side in shadow. The rig is very simple, I’ve used one area light for the main light source and a distant light behind his head to hint at some kind of rim lighting. The rest of the illumination is provided by the background Monte Carlo radiosity, set to 3 bounces using only a grey background colour.
The camera setup is also very basic, the focal length is set to 85mm roughly around the same as most portrait photographers use. Because I’m rendering it with Fprime i don’t need to worry about any of the AA settings as FPrime continually increases the quality of its AA with every pass that you leave it to do..
The image was rendered using FPrime, which is in my opinion the single greatest plugin ever invented, I can’t imagine working without it. Being able to render a scene in seconds with full radiosity, reflection, SSS and displacement is unbelievably helpful when your setting up a render, it allows you to tweak until your hearts content and not have to worry about lengthy low res preview renders every time you need to see an update. If your a lightwave user I cant recommend this plugin enough. For the bulk of the rendering I use the FPrime preview but when it comes to the final render I like to use the FPrime Render plugin as it allows you to set things like lighting quality and also seems to render slightly faster and with better antialiasing than the preview.
Above left: Lighting setup and properties. Above Right Camera Settings
Post Processing / Grade
OK so using the above lighting rig and render settings I rendered out the final image which looked something like this
Above Un-Processed Render from Lightwave (FPrime)
As you can see it bears very little resemblance to the final image its going to take a lot of post work to get it there, clicking on the image below will show you an animation of the various different layers used in photoshop and Digital fusion in order to achieve the final image.
Above left: Layers Video. Above Right: Photoshop layers
Levels / curves
I always use lots of adjustment layers, for this image I really wanted to add a lot of contrast so I used an ‘S’ curve and crushed the backs using the levels.
I did debate using Sasquatch for the hair however this image was only ever intended as an illustration and therefore I thought it would be a waste of time fiddling with hair setting and dynamics when I could easily paint it on in photoshop.
Depth of field effect
You might notice that in the render above his ears are out of focus, I actually created this effect very simply by using the blur tool in PhotoShop. Rendering DOF in LW can significantly increase the render time
chromatic aberration is caused by a lens having a different refractive index for different wavelengths of light. It creates a rainbow coloured fringe where the light splits into its component colours. I really like the effect as it adds an armature photographic quality to the image. you can see its effects in the image below I’ve exaggerated it quite a lot in this image. I created this effect in Fusion using the Speed six Monsters Plugin ‘Gun’
The grain was created very simply by creating a new layer in photoshop and applying some nice and setting the adjustment to ‘Colour Burn’.
Because I rendered the character out as a 32bit TGA I was able to use the alpha map to cut him out and insert a background behind. I used an image of a hospital emergency room and blurred it in Photoshop very simple stuff really.
The final step of the grading process took place in fusion using the ‘Speed 6 Monsters, Film effects’ plugin which gives you the option to choose from various different film stocks, I chose to under expose this image slightly, hence the greenish tint. you can see the effects in the image below
And that’s about it really, I hope this was of some use to you and if you have any questions feel free e-mail me firstname.lastname@example.org You can download the entire scene complete with all the maps and models from the link below.