This is one of my projects of my demoreel. The focus of those project is VFX Lighting and compositing. In this project I try to integrate a CG-Head into a live-action-plate.
· camera tracking
· optimate Skull Model (partly retopology, uvs)
· texturing and lookdev
· illustrate some 2D elements (logos, pictograms) for texturing
· model 3D geometry (sticker, some head applications)
As a VFX lighting artist you have to work with given life-action plates. So I wanted to do this project to show that I could handle this kind of workflow.
START / TRACKING A DARK FOOTAGE
After I found and selected the video footage and the model for the integration, I knew that tracking with this kind of footage would be a challenge for me. I liked it very much. But it was very dark, mostly washed out or blurred and had compression marks. This meant that tracking would be very difficult.
To get a better result in Nuke with the camera tracker, I sharpened the image slightly and increased the contrast between the person and the background without making the compression marks too dominant. To achieve an accurate tracking result, I chose a high number for the features and a low distribution of them. These adjustments helped me a lot with the final tracking result. However, minor errors in the tracking appeared in between. I manually compensated for these errors in Maya by adjusting the rotation of the head at the problem areas. In the best case you have tracker points on the person that improve the tracking of dark images.
COMING UP WITH A LITTLE STORY
After the tracking was done, I thought about how I could make the project even more interesting.
Since I had the model of a robot, I thought about doing something similar to what was done in the movies "Short Circuit" or "Chappie". The synthetic human should have got an own consciousness similarly as in the films by a disturbance. The concept was now the "free will".
For this I modeled in ZBrush a sticker for the forehead with the imprint "rejected". A graffity with an anarchy sign and the word "fake" should illustrate a conflict. Furthermore, I designed logos and symbols in Adobe Illustrator for the texturing process in Substance Painter.
MODIFICATE THE MODEL
To focus more on lighting, I downloaded the base model for free from scetchfab (credits in the demo reel). Since the entire model was made of triangles and not beveled, I decided to retopologize the larger elements (head, eyes) to have more control with maps later in the texturing process.
To make the model look more realistic and interesting in light and shadow, I added smaller applications to the model and worked a lot with displacement and bump maps.
LIGHTING IN MAYA
First, I added a simple gray shader to the model and imported it into Maya for reference. Then I roughly created the lighting. Without existing texturing I could see if there would be interesting shadows or if the model would look too simple at the end.
In the next step I created a cylinder for the first light and adjusted it to the size of the glow neon tube from the footage. After that I created 3 area lights and adjusted them to the position of the neon tube. The neon tube could now shine to all sides except the back. In this case, the back was not necessary because there was nothing to expose in the background. Since the surface was less shiny, 3 area lights were enough to give the impression of a glowing neon tube. More lights would have increased the render time unnecessarily at this point.
Once the area lights were in place, I was able to parent them to the cylinder and use it to control the animations. The cylinder has been rendered invisible so that it is not visible in any reflection.
In the next step I placed a skydome with a simple HDRI from an old warehouse. I adapted the HDRI to the existing colors of the background room with a color correct node. Now I was able to control the basic brightness of the room with the skydome I had created.
After all the lights were placed I built some geometry to use as a shadow blocker. Most important were the body and the collar. These have the function of blocking the light sources and at the same time were used to project the footage to give the head the right color cast.
After the lights were placed and ready, I prepared the required AOVs and set the render setup.
For control I checked again the set values for the Ray Depth in the render settings. To avoid unnecessary reflections, I went through the number of light reflections needed for the surface-comlexed objects in the scene. After some render tests with a small number of frames I was able to determine the optimal number of samples and lightsamples.
POST PROCESSING/ FINAL TOUCHES
After retouching out the visible hair with the Rotopaint node, bringing all the elements together was not particularly difficult. The trickiest part was to create the shadow of the head for the clothing using the Roto-node. For this I watched some videos and tested the light/shadow behavior of different surfaces like leather at home. Normally, the shadow is created using a shadow catcher. But because the tracking wasn't 100% accurate, I couldn't use the shadow catcher. Due to the effort involved, I decided to create the shadows differently.
The head felt more and more integrated when I adjusted the Z-Defocus setting to the footage. Next steps were color grading and adding some motionblur. Then I created a grain map and adjusted it to the footage to match the integration to the compression of the footage.
Final steps were adding godrays for the eyes, which I rendered from Maya, and using lensflare effects to make the eyes glow more. As a final step, I made the black values a bit darker to create a more menacing mood. I also used various vignettes to draw the viewer's focus a bit more to the head.
I learned a lot from the car project "Sunside ride" that I created before. I was able to apply a lot of what I learned there to speed up my workflow for this project.
The most difficult part of this project was the tracking, because in the best case you have tracker points on very dark footages, which simplify the workflow. From my current perspective, the footage may have been a little too heavy for me when I started the project. It's always good to allow a little more time for unexpected problems, but I learned a lot on this project about finding solutions away from my regular workflow.
Thanks to everyone who supported me during my time at Pixlvisn and Thinktank. Special thanks to Brandon Martin as well as Stefanie Knopp who provided me with feedback along this journey.
Also thank you for taking the time to read my entry. If you have any questions please feel free to contact me.
This project doesn't have any comments yet.