This is the third part of our freezing tutorial. We talk about integrating our elements in the scene and comping all together in theFoundry’s Nuke. If you haven’t seen it yet – this is our final result:

So far what we have is a virtual backdrop moving along with the keyed footage. Before adding new elements, we have to make sure to color correct and alter our elements (foreground element and backdrop) so they match up beautifully. Let us take you on an adventurous quest when approaching the grand finale of the final freeze effect.

Our goal is to merge our background with our foreground element. To do so we start by preping the cave background. Assuming you still remember the node tree from last time we continue working with the last scanline render node giving us the 2.5D render of our cave. Got it? Good. Since we are in fact in 3D space we can use this to comp all our stuff together. A ZBlur node could help, so lets try this. ZBlur actually works like you’d expect it to. It blurs objects in the distance more than those nearby. The scanline renderer still offers us depth information from our geo object so we dont have to set up anything special here. You can shift the focus plane to the point you need it to be, set the strength, then – like always – play around with the sliders until you’re satisfied, then go get some coffee. As soon as we stop staring at the ZBlur settings we could happen to notice something odd in the viewer. Like a bad bounding box. You can tell this by two dotted lines around your image where there should only be one. The difference between the lines is the actual shift between the footage (larger dotted line) and your viewer settings (smaller dotted line), meaning everything outside the viewer frame will not be rendered in the final output, unless you do something about it. Like using the reformat node. Just tell it what your output is going to be – root.format should be your first choice – and how you want your image to be positioned in the new frame. We chose center and no resize to just crop the image. So far so good. To give our backdrop a cooler look we turned down the saturation and tweaked the levels using a Grade node. We are facing a pretty common issue now: the keyed greenscreen footage has a very sharp edge and doesn’t seem to be influenced by the ambient light at all. The solution to our problem is the Lightwrap node. Connecting our color corrected backdrop to the B (as in background) input and our premultiplied protagonist as A input of the Lightwrap we are now able to control the fictive spill. See what it does to the edges? (In the image above we exaggerated of course). The edges blend better with the background now, but the sharpness of our actor’s outline is still too disenchanting. At this point we would usually create an EdgeBlur node to blur the edge a little and thus increase the impression of a small amount of depth of field. In this case we keep this step for later, as we are about to add elements – we would like them to be affected by the EdgeBlur too upon approaching the actor. So we’ll insert our elements first and then add the EdgeBlur. There is a big light hitting our actor from his right, so what we want to do next is adding a big flare coming in from the left side, accompanied by twinkling dust particles. The big flare is an easy to accomplish task: create a new RotoPaint node ( on your keyboard) and draw an organic shape on the left side. To create the actual flare just grab the vector points of your shape and this way extend the light by controlling the exact shape of the feathering. To blur it even more you can increase the feather property of the Shape inside the RotoPaint settings.

We did the dust particles inside After Effects, using CC Particle Systems. We won’t go into the details of particles as that’s not the overall topic, but using the Particle Systems settings below will get you the following effect:

We used pretty small, widely spread QuadPolygon Particles to see some depth, but you have to really tweak these settings to get something that works with your comp.

And this is what our dust and light part looks like after merging them together:

To implement all of these elements in our 3D comp we need to set up another quick 3D scene. Create a 3D Card, a Scene node and a Scanline Renderer. Connect the Merge node from before as the input for the 3D card and the rest according to this setup:

Note that we reused the cam used in our backdrop as a cam input of our new scene. This way we keep using the very same camera and easily preserve the same perspective. If you take a look at your composite you’ll discover there’s still one element missing in 3D: our talent. Yeah, we nearly forgot about him, but he’s starting to give us odd looks… Let’ss integrate him. If you recall our first post – Shooting and Keying – you can see all strings running together in a Copy node. The integration will start right here with a Premultiplication.

Since our dolly is not perfect we used a Stabilize node to remove some jittering. The Stabilize node is basically a CornerPin2D containing tracking data.

At this point you should be perfectly able to create this small 3D setup on your own, but we show you how it’s supposed to look like (for the sake of making the post a little longer, at least).

By pressing in the viewer you can switch to a 3D view. After zooming way out and panning a little you’ll get to see something similar to this:

That is our finished setup in 3D space. All there is left to do now is to do a final color correction and match the background. We start by merging the output of our Scanline Renderers together.

The youngest of you might still remember the EdgeBlur node mentioned earlier. It is time for us to make that sharp edge disappear. For once we take our color corrected and merged imagery (see Merge node in the image above) and use a Copy node (hit on your keyboard) to bring the alpha channel of our talent back in. Your result should have the actor with our backdrop and the light with its shimmering dust particles in the RGB channels but just the actor’s silhouette inside the alpha channel. Plug it into an EdgeBlur node and it will take the alpha channel as matte for its calculations.

We’ve gone a long way from shooting our footage, but we are still not at the end of our journey. Our last tutorial will reveal the magic behind the ice and snow elements.

See the result at the top of the page.

Jake & Dan

...you might also be interested in:

• Freeze Effect Part 2/4:
  The Backdrop
• Freeze Effect Part 1/4:
  Shooting and Keying
• Freeze Effect Part 4/4:
  The Snow and The Ice