I really enjoy still life, I think it is one of the best ways to learn composition and lighting. It has been a while since I’ve done a personal 3d artwork so I decided to refresh my memory and do some modeling, texturing and rendering. It took me about a week to finish the piece and I had a lot of fun making it. Since doing a bunch of apples and flowers would be a little bit boring I decided to create something creepy and moody.
Composition and Lighting.
I used a standard “L” composition which extends to a right triangle composition. The chiaroscuro lighting was my choice as I wanted a strong contrast between the light and shadow. In fact, I used only two small area lights, and to further emphasize the contrast I darkened the walls so they didn’t bounce the light on the objects. Willem Kalf’s paintings were the major source of inspiration.
Fast texturing for those who are lazy.
I had fun doing texturing work. I decided to use Photoshop for 3d texturing. While it is not as robust as dedicated 3d painting programs like Mudbox, Zbrush and Mari, PS can do decent 3d texturing and anyone who has little experience in texturing (like me) can quickly dive into the process using Photoshop’s standard tools. PS supports projections and direct 3d painting on a surface.
Physically accurate materials.
It can take a while to find accurate and appealing material settings which will look both aesthetically and physically pleasing. So I was curious to try the LollipopShaders Presets Library for RenderMan. The library has 100 material presets which match measured BRDFs of various materials from the MERL database. The presets gave me a good starting point and I didn’t have to worry about the physical accuracy of the settings.
The presets I used for my artwork.
Matching the reality to CGI.
In 99 percent of the cases it is the CG that needs to match the reality. Usually there is a live action plate and VFX artists do their magic to integrate the CGI. What about the other way around? Matching the reality to the CGI. That is what I was doing when I was adding the hands to the rendering. Actually they are not 3d, those are real hands which were holding a real plate under the exact camera angle and lighting conditions. To match the lighting and camera angle with the still life rendering I overlayed the rendered picture over a photo camera’s live view so I could see both the rendering and the real picture behind a lens. Then, I could easily line up every aspect of the reality to the rendered image. The overlay was done with the help of open source camera add-on Magic Lantern. To mimic the CG area light I used a tungsten 85 watt lamp bounced of a wall. The DSLR lens and camera settings were matched to the CG camera. Both cameras are Full Frame and use the 85mm lens. I even tried to simulate the Renderman’s noise, cranking the real camera sensitivity to high ISO.
Rendering and Compositing.
It took more than 10 hours to render a 3k image with new RIS Renderman. I’m not a big fan of long render times but RIS needs time to render a noise free image. Hopefully it will get faster in future releases. The compositing is always the most exciting part as it is the place where everything comes together. I split the image into different AOVs. One of the reasons I split the image was to perform noise reduction on certain shading components. For example, I could use the noise reduction on the indirect diffuse AOV without affecting the diffuse and reflection components. I prefered doing it in compositing rather than using Renderman’s denoiser as I could instantly see and control the result.
Once I was happy with the compositing I pushed the image through the “lensing” process which I described in my previous post. It gave the image a slight analog feel removing digital sharpness.
More to come. Follow @RenderStory on twitter to get all new updates on art, film, technology and animation.