Part 2 of our four-part tutorial covers the creation of our background and basic camera projection mapping.
The background we are going to be creating is the one used in our final shot:

To justify the dramatic lighting we were looking for a background where we could easily mix tungsten- and sunlight. We ended up with a picture found on sxc.hu:
As our original shot was a tracking shot, we knew we wouldn’t be able to reproduce a believable camera movement using 2D operations only. We needed the ruins in the background to be 3D.
With a big budget and endless, long nights with pizza and coffee we would’ve jumped right into 3D to remodel the whole ruins, texture and structure them with nice shaders and displacement maps and to fool around with render settings. Sigh… we went with the shortcut. If we can’t have 3D, we have to get by with 2.5D. In Maya we created a rough model of the ruins, which we ended up using for a camera projection of the original background inside Nuke.

Step one. Open up Maya and create a new camera. Name it projectionCam and go straight into your camera settings. Wait! Stop it! There is some very important information we are missing! In order to have a perfect camera projection, we need our camera in Maya to exactly match the very camera our background image was shot with. That’s why we went back to sxc.hu and took a closer look at the information provided with the picture. A small box on the right corner gives us the most important data we need: the camera he used.
Obviously we are victim of a typo here, the focal length must be 60 instead of 6mm. Also the aperture we are given is not giving us information about what the exact horizontal and vertical measurements of the aperture are. To be on the safe side and get as much information about the camera we can get we highly recommend www.dpreview.com. We went to Specifications > Olympus > Olympus SP-500 UZ (which, in our case, we had to find under the discontinued models) and looked at the Sensor size to determine our horizontal and vertical aperture. Back in Maya we copied our data from dpreview and pasted it into our projectionCam.
Once you’ve done this go further down inside the camera settings and go into the Environment settings to press the Create button.
The attributes for the created Image Plane pops up and we can hit the small folder icon next to Image Name to choose our cropped background. Make sure the camera options under View > Camera Settings are the same as ours above:
and change the render settings according to your source file to make everything match.
Now that everything was lined up we’ve built a rough lowpoly (very, very lowpoly) of the cave using planes and poly surface shapes for the arcs supporting the ceiling. We are not going into the details of modeling as this tutorial doesn’t focus on that, but know that it is crucial you build the cave walls using planes instead of a closed cube for the environment to work inside Nuke. It might be a good idea to blur and soften out the background-image temporarily (a layered PSD is a good choice) to make it easier creating the shapes on top of it.
Everything is lined up, but we are facing a different problem. Our cave is now bound to a replicate of the camera the background was shot with, whereas our footage is still shot with another camera. Both focal length and aperture won’t match, which would eventually make the compositing look odd or wrong. So we duplicated the projectionCam and renamed the duplicate projectionCam_footage. We changed the focal length and the aperture to the properties of our footage camera and looked through it in Maya. As position and orientation remain the same as in projectionCam, our image section didn’t change much. The whole foreshortening and distortion on the other hand was all messed up.
By rendering a quick garoud shading we had a template to retouch and distort the background-image inside Photoshop to match our new depiction of the 3d cave. Oh, and while we’re at it, lets throw in a little color correction as well. Colorize at will, try to match your desired style. As said before we wanted to create some intense lighting, so we darkened it and inserted some lights.

Back in Maya we exported our new camera as a *.fbx and the lowpoly as an *.obj file. With all these elements we go straight back into nuke.

This is what our projection setup will look like:

Let’s walk you through this node tree step-by-step. We start by creating a Read node and load our previously generated and perspectivly corrected cave image. Because we do not have an alpha channel in this image although working in rgba colorspace we need to create one really quick by running our image through a Shuffle node, setting white as our alpha channel.

After premultiplying it we can create a Project3D node followed by a ReadGeo node with our *.obj file loaded into it. Remember the camera we exported earlier? Right, we need to implement it as well, so create a Camera node, check the Read from file checkbox, go to the file tab and select your *.fbx file, which should contain the camera information for a good and solid 3D projection.

By the way, you can change the background color of your nodes by clicking the button right next to the nodes name. This is just so you won’t get confused with different cameras in one scene.

We now have all our 3D elements set up very quickly so let’s connect them in 3D space. To do so we create a Scene node which holds all our 3D elements together. Connect the Projection Camera (projCam) and the ReadGeo node to the scene and output it to a Scanline Renderer. If you look through the scanline renderer you shouldn’t be surprised by only seeing your plain image. You have to switch to the 3D view (by pressing inside the viewer panel) to see something more interesting:

If you think we’re done by now, we have to disappoint you. There is one more step to take: the movement in 3D space. You can approach this from various angles.

Tracking
Track your original footage and load your 3D tracking data into a new camera

3D Animation
Animate your camera in Maya, Cinema 4D, 3D Studio Max or your 3D programm of choice, then export it and create a new camera with your animation in Nuke.

2.5D Animation
Animate a newly created camera in Nuke, which was what we did. We could have used any of the other methods as well, but we wanted to have more control over it. You’ll learn more about the integration of the background and the footage in part 3 of our tutorial, so stay interested!

Whatever technique you used, just connect the moving camera to the Scene node and as cam input for the Scanline Renderer. NOW we are done. Stay tuned for the third part of the Freeze Effect Tutorial!

Jake & Dan

...you might also be interested in:

• Freeze Effect Part 3/4:
  Integrating The Elements
• Freeze Effect Part 4/4:
  The Snow and The Ice
• Freeze Effect Part 1/4:
  Shooting and Keying