Frankenduck Image Composition
Project 2 of 2 for CSANM 354: Shader Programming, completed in December 2020.
If the first half of my Shader Programming class at BYU was focused on recreating a real-world object, then the second half was focused on trying to seamlessly blend that object into a photograph. The goal for this project was to put together an image with a fake object so that the viewer wouldn't be able to tell if it was real or not.
Step one was to take a picture of some objects with the intention of removing one and replacing it with a CG version. I love Halloween and had a lot of little decorations lying around at the time, so I chose to set up a little group of them on top of my dresser.
Step two was to actually recreate the object in the computer. I chose the Frankenstein duck near the center of the image, and modeled the basic shape of it with Autodesk Maya to start. All the textures were done in Adobe Substance Painter, including the displacement maps for the wings and the bolts on the sides of its head. Special care was taken to recreate the duck as closely as possible; the green, black, and orange paint was much shinier than the matte grey plastic the duck was made of, and there are places where the paint doesn't exactly match the molding of the body, such as the hairline and the collar. The original duck was imperfect, so my digital recreation was intentionally imperfect as well.
Step three was to take the finished model into Houdini and set up a very simple environment to render it out with. A virtual camera was made to imitate the settings of the real-world camera that took the photo, simple versions of the dresser and one of the decorations were made to-scale so that lining up the shot would be easier, and then the duck was put in place. From there, the dresser and block were made invisible and the duck was rendered all by itself, then the duck was made invisible and a shadow was rendered.
The final step was to take all three parts of the image into Nuke and composite them together. One thing the Houdini camera didn't replicate was the slight blur and grain of the original photo, so those had to be added in during the compositing process.
In order, we have the image without the object I chose to recreate, then the model without any shaders superimposed onto the shot, the model with the Substance Painter shaders, and finally the finished image with the shadow map and the slight amounts of blur and grain necessary to help the object blend in with the original picture.
Comments (0)
This project doesn't have any comments yet.