A Sword Nobody Asked For: LSU Filmmaker John Chua on Making Movies with AI

John Chua is an assistant professor of film and television in the LSU School of Theatre & Film. Before coming to LSU, he spent nearly two decades teaching abroad, more than a decade of it in London, while writing, producing, and directing documentaries, animated features, and a thriller. His films were distributed by the BBC, PBS, Sony, and Universal, among others. Chua’s newest project is a two-minute animated short made with the help of AI, selected to screen at Los Angeles Tech Week’s Neu Wave AI Fest on October 14. He spoke with LSU Research about how the film was made, what went wrong along the way, and why he thinks “AI” has become too big a concept.

It’s a two-minute animated short with the feel of Japanese anime. I was experimenting to see what I could do—using it to learn. It’s just one of my experiments—not a masterpiece. It’s the third thing I ever made with AI, for goodness’ sake. People are going to like it, or hate it, or be indifferent. If they find out it’s AI, some might say it’s crap without even thinking about how it was made.

I was testing the software, a platform called ElevenLabs. It’s popular in the media world—Disney has backed it through its accelerator program—and I was able to get us to be their Impact partner, so they gave me and my students free access. The software is fantastic. It does many things: audio, visual, everything. As a faculty member, I need to be up to date on what's actually happening in the industry.

ElevenLabs is a platform, and within it you use different models. I use Google’s Nano Banana to create a beginning image and an end image—what we call keyframes. Then I use Seedance or Veo to move an object from A to B. You tell it, “The person is going to run from A to B,” or “walk slowly from A to B,” and it will do that as long as you have the beginning and the end. In traditional animation, someone has to draw each frame of the person walking. Here, the computer is doing the drawing. It helped that I’d worked in animation before, as a writer, so I knew how it would work, and I tried to push the software to its limits.

This two-minute short took me 40 to 45 hours, which shocks people. AI is not perfect. It keeps producing mistakes, you keep correcting them, and it takes a long time to edit. When you go online and see stuff that’s obviously AI and badly made, that’s because they churned it out very quickly without correcting the mistakes. It’s the equivalent of reading a freshman essay straight out of ChatGPT and saying that’s the standard of what AI can produce. Just because the vast majority of what you see is bad doesn’t mean it represents what AI could do. It’s a misconception that it’s all just “AI slop,” and I’m trying to correct that.

AI often misunderstands the prompt and produces extra stuff that’s not in the prompt. In my film, one of the characters suddenly pulls out a sword. It’s not in the prompt, which was something more like “She’s waving wildly.” I don’t know where the sword came from, but the AI result is based on a million other images out there and trying to approximate something. That’s one of the weaknesses of AI, but I think it will get better. That’s why there are events like the one I’ve been invited to in Los Angeles, where investors and venture capitalists are trying to find solutions to exactly these kinds of problems.

I wouldn’t have been able to make it at all. Animation is frame-by-frame drawing. Traditional hand-drawn animation would take weeks, maybe months, and I don’t have that kind of time.

We hope to plan a conference later this year to understand the opinions of our community. Many people see AI as a negative development in the arts. I don’t like forcing students to use AI; I prefer just demonstrating AI and they can test the tools because students need to know how they work since AI is used extensively in the industry. They need to be aware of these tools and what other professionals are already doing, and make their own ethical decisions regarding using them or not.

The term is too big. AI is so many different things, not just ChatGPT. I teach students how to use a camera, and the camera has AI built in: if the subject is moving, you tell it to track the face, and if a hand comes out of nowhere, it won't jump to the hand—it sticks to the face. It’s analogous to how my car detects a pedestrian or a bicyclist and slows down. It’s been trained to look for humans. AI is incorporated into a million things. Many of the Hollywood films and TV shows you see have some kind of AI in them. They just don’t tell you, and people don’t notice. There’s a Netflix series about Brazilian soccer, Brasil 70, where the stadium crowds in the background are all AI—and nobody cares. But if you make something entirely in AI, it’s going to be noticeable. I am not inclined to make a lengthy film entirely through AI generation because of its current limitations.

I was there to promote U.S.-Ukraine relations, as a program of America House, the cultural outreach arm of the U.S. Embassy. One of my friends was the highest-ranking diplomat at the U.S. Embassy in Kyiv at the beginning of the Russian invasion, and later ambassador to Armenia, so I knew her. I was there to show AI is both a dangerous and useful tool. It can be used for propaganda, and there are many, many ethical issues with AI filmmaking. We’re only starting to grapple with them.