Compositing in the Age of AI-Enabled VFX

It’s easier than ever to create just about anything you can imagine today. Traditional barriers no longer limit us, and generative AI models can output stills or video in minutes or seconds. While these tools pose new opportunities, artists should continue leading with technical knowledge and techniques, rooted in traditional visual effects workflows to protect their creative vision. 

In content creation, compositing is where the visual ingredients become fully baked scenes. Human artistry is essential at this step. AI can be extremely beneficial in many aspects of content production, but over reliance on the technology can skew visuals into reading fake.      

At Foundry we understand the production realities and pain points creatives face and let it guide our development. We build tools that empower artists — from ideation to final output — by accelerating tasks and enabling new workflows, while retaining complete creative control. 

I recently spoke with creative technologists Ben Abergel and Freddy Chavez Olmos about their experience with AI in production, how the technology has impacted their approach to compositing, and where they think the industry is headed next.

The University of Southern California (USC) Entertainment Technology Center (ETC) is driving filmmaking forward through exploration and experimentation with emerging technologies. Its work has led to groundbreaking virtual and cloud-based productions. Currently, the center is working on a pair of AI-assisted short films, with Abergel serving as the Head of AI Technology and Production for the hybrid live action production project.  

Embracing many traditional VFX processes, the production team films actors on a green screen stage, brings tracking data from those shoots into a virtual environment, and uses AI to retexture the background. Additionally, they’re leveraging AI to help extract all the VFX passes and assist with relighting. Deploying AI on projects like these allows for a small production team to move faster, and then they can refine outputs in compositing. In Nuke, the team retextures backgrounds or retracks on-set footage if the captured data is imperfect. It’s also helpful for adjusting scene parallax.  

For Abergel, his background in traditional VFX, such as rotoscoping and compositing, has been essential to working with AI. He says planning has also been key to achieving the desired result faster, though he thinks there’s much to learn by embracing the chaos that comes with AI. Abergel believes that AI acceleration will free up time that could be reallocated to pushing the creative envelope and bring teams together earlier for improved efficiency.

Above all, he stresses that maintaining artistic control is non-negotiable, and notes that control is broader than ever. Artists are no longer working on one small element of a shot, but rather, the entire shot so they need to ensure everything ties together. This often requires connecting disparate pieces, despite tool limitations. In this regard, problem-solving is a vital skill for AI-supported production. 

Olmos, a multidisciplinary creative, filmmaker, and VFX artist, was first intrigued by AI’s VFX potential with the release of Beeble’s Switchlight model, which generates physically based rendering (PBR) passes from video. Implementing it on the television series Superman & Lois, he was able to speed up the relighting process and give the production’s compositors a tool for creative, interactive lighting and other effects that would have been too time-consuming or expensive to execute otherwise. 

In his view, the fundamentals of VFX and filmmaking are imperative today, as the industry moves toward more generalists and storytellers with the rise of AI-enabled workflows. Having begun his career as a makeup effects artist, he carried that unique knowledge of human anatomy and texturing through to his VFX work. Similarly, understanding core aspects like lighting, composition, and timing has helped him generate more useful AI outputs. While compositing has historically been a later stage step of the pipeline, he’s seen workflows become less linear and compositing evolve into a full creative control hub, where teams integrate and render CG elements from other departments. In particular, he’s found being able to relight plates in compositing to be incredibly valuable.  

Despite his enthusiasm for AI’s myriad possibilities, Olmos cautions that the technology is still limited. Clients can overestimate AI-powered capabilities and need a stronger understanding about what is realistic for productions today, and what’s more future looking. AI tools have also reshaped creative iteration. Often Olmos will zero in on what a client wants by process of elimination. This helps push creative decisions forward and minimizes big post fixes. 

AI is undoubtedly impacting where VFX is heading, and quickly. It seems like there’s something new to explore almost daily, which is very exciting, but the chaos can be daunting. We’ve gone from hand-painted models and cel animation to digital workflows that support ever-increasing resolutions and higher dynamic range. Each advancement, no matter how large or small, expands the creative toolbox. For example, Foundry’s February 2026 acquisition of AI orchestration platform Griptape will make it easier for studios to leverage AI models in production without disrupting existing workflows. 

The AI genie is out of the bottle. Now, it’s a matter of figuring out how we move forward ethically and effectively, and above all, preserve strong storytelling, regardless of how those stories are realized. We’ll soon move past conversations about efficiency and technical limitations and refocus on creating amazing visuals that inspire audiences.