---
title: "Stirling Compositing"
canonical: "https://rmanwiki-27.pixar.com/space/RU/657162271/Stirling%20Compositing"
format: markdown
---
### Compositing

That was quite a journey, but we’ve finally reached the last step of our integration: compositing!

First, we’ll ensure that our ACES workflow remains consistent in Nuke as well. No worries—this only takes a few clicks and Nuke will convert the sRGB footage into proper acesCG color space for us.

![image3.jpg](media://304524c1-171f-4374-902c-b29ca334d201)

To access your project settings in Nuke, hover your mouse over the Node Graph window and press "S". Navigate to the Color tab and switch the "Nuke Default" setting to "OCIO". This ensures that Nuke automatically opens all your files in the correct color space.

*Now, feel free to download the project file, where I’ve added plenty of sticky notes explaining the entire compositing process node by node.*

One useful Nuke workflow tip that helped me avoid draining and time-consuming rotoscoping is the use of 3D cards to generate additional masks. 

Here’s how the setup works:

![image18.jpg](media://48486fc9-cd10-4a3f-b3d3-6f3de3ed0988)

When writing your image sequence of the generated mask, make sure your .png will be generated with an alpha for better results.

![image14.jpg](media://23cff333-0ffd-444e-be10-ac7d5fcbbb4d)

This basic 3D scene setup, combined with a dual-viewer workspace, allows you to draw “directly” on flat, non-moving surfaces—such as the ground. By creating a single rotoscope shape and using the camera tracking data, you ensure that the rotoscope remains stable throughout the shot, saving you a significant amount of time and effort.

In this case, I used this technique to generate an occlusion mask for the shadow our car will cast in this part of the footage.

However, if you find yourself with no other option but to tackle the dreaded task of manual rotoscoping, arm yourself with patience, determination—and perhaps a good podcast to keep you company.

![image5.jpg](media://34c979bb-25ae-42f2-8353-23adc04c92ca)

When it comes to rotoscoping, my best advice is to keep your shapes as simple as possible. It’s much easier to work with 20 smaller shapes made of 6 points each than to struggle with a single, overly complex shape that will likely drive you insane by the end of the day. So, don’t hesitate to break down your object into multiple shapes for better control and efficiency.

Another useful trick is adjusting the "Life" input for each shape. This allows you to deactivate a shape once it’s no longer needed, preventing unnecessary shapes from awkwardly floating across your screen.

![image17.jpg](media://14e42e6a-0bf1-44f7-aa3f-31fc9bffdb9e)

Additionally, fewer keyframes mean fewer problems. To save time, start with a rough pass by setting a keyframe every 5 to 10 frames (you can use Shift + arrow keys to jump by a set number of frames). This will help you quickly identify which frames require the most refinement.  
Finally, while Bezier curves are the go-to tool for rotoscoping, I highly recommend experimenting with B-spline shapes. They often require fewer control points, making them easier to fine-tune and stabilize, especially when working with organic motion.

![image13.jpg](media://cd7cd5af-1d00-475f-b92b-563313d6d3b1)


Feel free to explore the provided Nuke scene to get a deeper understanding of the entire process. You’ll find sticky notes explaining the node-by-node process.


However, if I had to summarize the key things to watch out for when working on plate matching, based on my own experience:

- **Extreme values:** Anything too dark or too bright will instantly break the illusion. Keep an eye on contrast and ensure smooth integration.
- **Pixel imperfections:** 3D renders tend to be too sharp straight out of the render engine. Take the time to soften edges and blend them naturally into the imperfect real-world footage.
- **The illusion matters most:** In some cases, I even rotopainted out small details—even if they were real—because they were too eye-catching and could make the viewer subconsciously feel that something was off.


At the end of the day, the goal is to seamlessly blend CGI into live-action while avoiding anything that might trigger visual suspicion.

I personally see this as a very unforgiving, but creatively exciting exercise, because even the most inexperienced viewer can instantly tell when something feels off. Our eyes have been naturally trained to observe the world around us for an entire lifetime, making any inconsistency in the integration immediately noticeable.

Have fun with the project!