CG Head Integration Project

For a job interview recently, I was tasked with integrating a CG render into a photograph. I was provided a Maya scene, set up with shaders and a model of a female head. I then had to light, render, and composite the CG into the original image. I was asked to complete the task in around 6 hours, and it ended up taking me around 7 hours. The goal was to make the final result as realistic as possible so that a casual viewer couldn’t tell the image was CG. Overall, I am happy with my results and it was actually enjoyable to work on.

Lighting and Rendering: Maya

The texture and model that I received were high fidelity. Even before setting up proper lighting, early renders produced nice results. For rendering I used Arnold since the scene file I was provided was already set up with Arnold shaders. However, even if that wasn’t the case I would had used Arnold anyways, since it is well suited for photorealistic renders. A downside of using Arnold is that renders can take awhile, however this scene was small and render times fast so it wasn’t an issue. Most renders took around a minute, but for the final render, once I was happy with the look, I cranked the settings up and did a 30 minute render. There wasn’t much difference between the renders, but since this project only required a single frame waiting a half hour for a frame wasn’t a big deal.

The lighting set up for this scene was fairly simple. All I used was a environment light and a directional light. For the environment light I choose a late afternoon HDRI scene with strong orange and red tones to match the skin tones in the original image. The directional light’s main purpose is to recreate the nose shadow. I choose a directional light instead of a point light because the nose shadow in the original image is hard where a point light would result in a soft shadow. I also set the directional light to a warm tone. To create the shadows around her chin and screen left ear I used light blockers on the environment and directional lights and adjusted the opacity for varying shadow tones. I also adjusted the shadow color to have more blue cool tones.

Compositing: Nuke and Photoshop

I wanted my final image to have strong specular highlights, especially around the nose. I felt that was a strong characteristic of the original image. In the raw render the specular is really soft so in the composite I turned the gain up high on the “coat” layer to give her more highlights. I also color corrected her lips more red and eyebrows darker. I felt her eyes were too red so I desaturated them, and I wanted the pupils to be more visible so I made them a lighter shade of brown. Then I increased the specular in the eyes to make them fell more wet and glossy. Below you can see some iterations of these changes as I was playing around with the final look.

Through out this iterative process I was going back and forth between Nuke and Maya making tweaks to both the render and composite depending on the issue. A challenging part of the composite was integrating the shadow around her chin and neck. I tried to get the shadow color close in the render, however I didn’t want to use a really saturated light shadow color in Maya. I felt that would be unrealistic and could lead to unusual issues later. In the end, I used a roto shape to pull some saturation out of the shadows and tweak them to match the shadow on her hand. Also, I created a roto shape shadow for her fingers since that would have been to difficult to match in Maya without using a model of a hand.

I did a version of the final composite both in Nuke and Photoshop since the interviewer wanted a version in Photoshop. However, since I feel more comfortable doing quick iterations inside of Nuke I did the major of look-dev in Nuke. Once, I had a final image I liked I then recreated it in Photoshop. Below you can see side by side the final image from Nuke and the final image from Photoshop.

I think the final result turned out well, hopefully the company I am interviewing with thinks so too.