Expert tips for perfect CG portraits

Past 3D World cover artist Francisco A Cortina presents his five killer tips for improving your 3D portrait images

Francisco A Cortina has been involved with the 3D graphics industry since 1995, and has worked at the highest level on feature films, videogames and animated shorts, including Final Fantasy IX, Shark Tale and Aeon Flux. Yet his roots lie in fine art: a grounding that has informed his work throughout his career.

“My biggest inspiration has always been fine art, specifically painting, because of its unique ability to manipulate the effect of light on canvas,” he says. “Even though I’ve spent little time painting in the last few years, the artists that still inspire me most to this day are Dalí, Degas, Ingres and Alma-Tadema.”

We asked Francisco to send us his five tips for better portrait renders. The results provide a fascinating insight into his working practices.

01 REFERENCE

.

.
Before beginning any actual work, I collect as many reference images of my subject as possible. If there’s a chance to make use of a photo shoot, I pare the results down to a set of eight hero photos, starting from the front of the face then circling around at intervals of 45 degrees. I’m careful to select images with similar focal lengths, sharpness and exposure.

During the early stages of development, I create a camera [within Maya] used only for reference work. On that camera, I attach a background image plane that dynamically loads each of the reference images per frame. To aid in aligning the model to each of the images, I keyframe the camera at every frame. That way, whenever the Time Slider is changed, the view automatically changes to match the reference image behind it.

02 MODELLING

.

.
The modelling phase is a careful balancing of technical construction and artistic sculpting. When building a model from scratch, I establish the structure of the base model in my main modelling package, Maya, then do a series of passes in either ZBrush or Mudbox, exchanging the model back and forth between the applications.

These days, I use as many edge loops as possible, and avoid n-gons and triangles. As the eyes are key to capturing someone’s personality, I always focus there first, working my way outwards, keeping in mind the proportions of, and distances between, the nose, lips and eyes.

When laying out UVs on the face, I put the texture seams where the hair parts on the top of the head. Overall, I’ve found it better to have minor distortion in a few areas than to have too many UV shells, especially when using 32-bit displacement maps.

03 TEXTURING

.

.
No matter what software I’m using, I preview and paint textures with a flat surface shader and no diffuse shading. Photoshop Layer Comps are the best way to manage variations of texture maps. I label each one with the corresponding version of the scene file and date in case it needs to be restored later.

Since some applications don’t preserve Photoshop Adjustment Layers or Layer Comps, I keep a separate painting PSD and a master PSD, with all of the integrated paint work and versions. The colour map should have no baked-in shading, odd highlights or reflections, to ensure that it reacts properly to lighting. The specular map and any control maps for the Fresnel and Fresnel Edge intensities should coincide with the skin’s colour and bump layers.

04 LIGHTING AND RENDERING

.

.
Lighting can make or break the look of a character’s face. In fact, it’s one of the last big challenges for CG. Although we’re still some time away from a true physically based CG technology for simulating the interaction of light with human skin, it is possible to fool our audiences into believing a CG face is real.

When lighting my scenes, I use a combination of HDR dome lighting with traditional key and rim spot lights. Final Gathering is my preferred render mode: I usually set the Intensity of the dome to about 0.3 and the key light Intensity to around 0.8.

Sometimes I even place ‘cards’ [planar objects positioned around the model to reflect light onto it]. These have a flat surface shader with Incandescence turned on: this helps to achieve good eye reflections (that’s where you can win or lose a viewer) and simulate natural real-world lighting.

05 SHADERS

.

.
Before getting to the bells and whistles, I make sure my basic shader channels are working properly by test-rendering them individually. When working in ZBrush or Mudbox to create the bump or displacement passes, I keep the general sculptural displacements in a separate layer from the fine pore-level details.

I usually postpone the very fine level of detail sculpting until the general colour map is done, so that the small wrinkles, grooves and blemishes can be matched up properly. In Maya’s mental ray subsurface skin shader (misss_fast_skin), I map a modified version of the colour map in the epidermal and subdermal layers.

I also map Reflect Weight and Reflect Edge Weight to a combination of the Bump and Specular maps, mixed with some Gaussian Noise. For the iris and lachrymal gland shaders, my Subsurface Scattering colour is pure white, and about half the intensity of the one used on the skin.

For the reflective parts of the eye, like the tears, conjunctiva and tear catcher surfaces, I have a Phong shader with a mostly transparent and reflective sheen, controlled by a Surface Sampling node (mapped to a Ramp’s outV) to create a fall-off perpendicular to the camera view.