Peter Ganunis ’25, a dual film major in the College of Visual and Performing Arts and computer science major in the College of Engineering and Computer Science, has set his sights on bringing filmmaking artificial intelligence tools up to professional standards.
For years, Ganunis has been incorporating artificial intelligence into his artistic process, even writing programs to generate music. He is leveraging these tools to complete his thesis film, a feature-length, 3D-animated musical.
“The way I’ve always thought about making art is about the process,” Ganunis says. “When I’m taking something on, whatever the medium, there’s one end goal of what I want it to feel like. How do I trick myself and trick the tools to get to that point?”
As image and video artificial intelligence tools gained traction, he received funding from the Syracuse Office of Undergraduate Research & Creative Engagement (SOURCE) to work on a style transfer algorithm.
Style transfer occurs when content is altered to match a specific style set by a reference image, such as changing a hard shadow to a brushstroke. Traditional tools often produce flickering films with changing colors as the video moves from frame to frame, Ganunis says.
He developed an algorithm to address this issue that involves rendering, or transforming the 3D virtual space into 2D images, two times. The first rendering creates images with recognizable objects like trees.
The second rendering produces a chaotic blend of colors and noise that the computer is able to track. When the computer finds two frames to be similar, it copies the stylistic changes it made in one frame to the next, adding a level of consistency to the process.
Ganunis is still working on the 3D animation for the film. He has been using a Rokoko motion capture suit, acquired with funding from the SOURCE, to capture character movements with help from his roommate, Julian Nowik ’25, a film major.
“Once we’re finished with the 3D animation part, we’ll be able to just render it out as the 2D-animated style and go from there,” Ganunis says. “The end goal is to get this old- school, 2D-animated Studio Ghibli look.”
Ganunis also developed an algorithm to aid in his work as a colorist. The tool helps change colors in an original image to match a reference, allowing filmmakers to achieve specific looks, like the dark blue color palette of a classic horror movie.
While Ganunis plans to make his work on style transfer accessible online after his movie is released, he intends to work on integrating the color algorithm into normal video editing software for use as a plugin.
Ganunis hopes to continue combining his love of art and computer science after graduation.
“I’m looking at a couple positions I could take in animation and a couple in computer science, but really what I’m interested in is the intersection of those two, doing programming work for image processing,” he says. “I’m excited to try to find out where my place in the world could be within image processing.”