Chapman senior creates fully AI-generated thesis film project

Photo Courtesy of Betsy Siegel

“Fully, entirely (artificial intelligence) AI films are pretty rare,” senior creative producing major Betsy Siegel told The Panther. “In the entertainment space, normally you will see some sort of human engagement with it. Typically through some form of voice work or music or writing. So, it's really uncommon to have AI do every single step of the way.” 

For creative producing seniors in Dodge College of Film and Media Arts at Chapman, their senior thesis project sees them budgeting, scheduling, permitting and recruiting a crew to create a 15-minute short film shot either on location or in the sound stages at Dodge. For Siegel, the process was entirely different. Instead, she employed AI programs that generate text, audio, image and video to demonstrate the potential of these technologies. 

“I was aware that the school was trying to implement AI into the program,” said Siegel. “I told them, ‘Hey, I know you really want to do this. I’m speaking at these AI film festivals all the time and I know that there’s a real place for this.’ And so I pitched them, ‘Would you want me to do an AI thesis so that you can then show it to people and be like, ‘Hey, here’s what our students can do with this?’’ So now I can go through and I can test out these other forms of software that you’re really interested in knowing more about.’”

For Siegel, working professionally with generative AI was a no-brainer. Her father worked in cyber security and early AI development throughout her childhood, meaning dinner table conversations about AI were not uncommon. Her mother, a writer, eventually got a masters degree in artificial intelligence ethics, and the combination of her parents’ disciplines inspired Siegel to pursue a law degree specializing in AI entertainment law. Having worked as a freelance AI consultant, Siegel’s background served her work.

“Prompt engineering is the language that you speak to AI with,” she told The Panther. “Because I came from a tech background, I knew how all these different things worked and I understood the data behind it. Because I came from a literary background, I figured out really quickly how to phrase things to get the best product possible.”

She continued: “I started helping people around me get similar products because they took longer figuring out how to properly craft prompts. I learned this one tiny niche thing and it just kind of snowballed, and it was the right time because the demand was about to explode.”

Siegel started the process of creating her thesis film by asking ChatGPT to write her a poem, which would be narrated by a protagonist. With the poem serving as a script, she proceeded to Pinterest to find reference images of the visual style she wanted for the short film. After plugging these images into Dream Studio, one of the leading AI image generation softwares, Siegel created a new set of AI-generated images inspired by the references. The next step was turning these images into video, for which Siegel utilized LTX Studio. This program expanded each image into a two to five second clip, created a consistent design for the protagonist, suggested new ideas for shots and organized a preliminary editing timeline that Siegel worked off of. 

For the protagonist’s narration, Siegel recorded her own voice and modified it with Kits Voice Studio to emulate an original voice for her lead character. As time went by, the programs improved drastically, and after months of regenerating clips and reediting them, Siegel realized the process could turn into a never-ending cycle. 

"I had to redo so many shots just because the technology was improving as I was making the film. Some of the shots that I made in the beginning are noticeably lower quality compared to the ones I made more recently,” Siegel said. “It kind of came to the point where I realized I had to stop reediting it because the technology is just going to keep improving, and I do at some point have to turn something in.” 

Siegel continued: “You could literally do this forever because the technology is constantly updating. I just had to accept that this version of the project is more of a time capsule than a final, perfect product."

Siegel sees potential for this technology being used and taught in the classroom, as well as harnessed in creative industries for particular purposes.  

Photo Courtesy of Betsy Siegel

“I think that image generation should be used a lot more regularly,” said Siegel. “As a producer, in almost every class I've had, we've had to make a pitch deck. These are decks that we need to be able to put in our portfolio and carry out into the world so they need to look good and they need to look professional. I think that there should be part of a class that covers how to make pitch deck designs with AI.”

She continued: “I think if people want to use this for previsualization, that's great. It helps to go into a pitch with, ‘Here are the AI images we created that are essentially what we want visually when we shoot — we're going to emulate these things.’ I think that using these as the foundation to then creating something that is made with people is the best way to go about it, instead of just replacing the people. Getting rid of production is a mistake. I think that these are just really good building blocks to then learn how to do the harder stuff on top of it.”

Artificial intelligence-driven film production is rapidly expanding and being invested in from many corners of the film industry. Staircase Studios AI, a new company founded by the “Divergent” series producer Pouya Shahbazian, claims they are making over 30 studio-quality projects in the next four years, each with a miniscule budget of less than half-a-million dollars. “South Park” creators Trey Parker and Matt Stone’s deepfake startup Deep Voodoo raised $20 million from a venture capital partnership led by Creative Artists Agency, while “Ready Player One” star Tye Sheridan’s AI-powered visual effects startup Wonder Dynamics raised $9 million from Samsung, Epic Games and others. 

Major studios are also pivoting in the same direction; Disney recently established an Office of Technology Enablement, which aims to integrate AI and mixed reality (XR) into Disney’s future creative projects. 

Stigmas have developed around generative AI in creative communities such as film schools due to concerns regarding plagiarism, labor replacement and other ethical issues. Among rapidly changing conditions, Siegel has a positive outlook on creative employment opportunities and insists that those that become well-versed in these evolving technologies will be better prepared for what the future holds. 

“My favorite analogy for this is a carriage wheel factory. When cars were invented, people were all afraid that the jobs would go away in the carriage wheel factory because they wouldn't need carriage wheels anymore,” she said. “And now it's a car wheel factory. The jobs didn't go away. They just pivoted. You might not be making the same thing, but there's still jobs available.”

She continued: “I know some of my peers who are very resistant to AI are just afraid. There's no need to be afraid that there's not going to be any job opportunities. It's just that the jobs are going to be a little bit different than what you thought they were going to be.”

Siegel’s thesis film will be screening at Marion Knott Studios on 5/16. Admission is open to all.