The Washington PostDemocracy Dies in Darkness

AI can make movies, edit actors, fake voices. Hollywood isn’t ready.

A new short film made with AI highlights the possibilities and pitfalls of using the technology in entertainment

Chad Nelson, the director and writer of "Critterz,” animates the face of an AI-generated character on his computer screen. Nelson used OpenAI’s DALL-E to create his short film. (Courtesy of Chad Nelson/Native Foreign)
7 min

It took Chad Nelson about a week to make thousands of photos of furry creatures and magical forests using Dall-E, an artificial intelligence image generator that has gone viral over the past year. Now, he’s made the first animated short film that uses images he generated exclusively from the AI tool.

Nelson’s five-minute film, called “Critterz,” was released online this week and introduces viewers to cuddly creatures that inhabit an imaginary jungle, resembling a cross between a Pixar creation and a David Attenborough-style documentary.

It’s an early example highlighting the possibilities and pitfalls of using artificial intelligence in filmmaking, a development that both excites and worries Hollywood.

Nelson, a visual artist in San Francisco, didn’t rely on AI for the entire production: He wrote the script himself, and enlisted actors to record the audio and animators to bring the creatures to life. It would normally take a large staff six full months to create the kinds of high-quality images in “Critterz,” its movie director Nelson said. But using OpenAI’s Dall-E, the process went much faster.

“It definitely took a lot less time, and it took a lot less money, than if I had done it the traditional way,” Nelson said in an interview with The Washington Post. He experimented heavily with Dall-E, using prompts such as ones to make “a cute striped fuzzy monster with small horns peeking over a mossy hill in a misty forest, backlit” to quickly create his characters. OpenAI, the San Francisco AI lab that created Dall-E, helped finance Nelson’s film.

Independent movie makers and Hollywood studios have been early adopters of generative artificial intelligence tools, which can create text, images and videos based on troves of data. These increasingly advanced products can save time and resources, their proponents say. It’s making Harrison Ford look younger for his upcoming “Indiana Jones” film. It gave Val Kilmer his voice back for “Top Gun: Maverick.” It made Thanos more closely resemble Josh Brolin in “Avengers: Infinity War.”

But the entrance of these tools is causing trepidation. A Goldman Sachs report in late March said generative AI could significantly disrupt the global economy and subject 300 million jobs to automation. The Writers Guild of America, which represents screenwriters, is locked in negotiations with movie studios — and the way artificial intelligence can be used in scriptwriting is a key sticking point. Actors, such as Keanu Reeves, are raising alarm bells, saying the rise of generative AI is “scary” and could be a way for executives to not pay artists fairly.

How widely AI is adopted in Hollywood hinges, in part, on how broader issues of intellectual property, consent and contract negotiations play out, lawyers and media experts say.

“Generative AI is really a game changer,” said Ryan Meyer, a copyright expert and of counsel lawyer at Dorsey & Whitney, but “there’s a lot of issues … that need to be resolved.”

Hollywood is no stranger to artificial intelligence. Before the recent rise of AI chatbots, image generators and voice modifiers, studios used the technology to fill out battle scenes and for digital animation, said Joshua Glick, a film and media studies scholar at Bard College.

But the images, text and audio that generative artificial intelligence companies can create now are far more realistic, he said. They’re not just tools that professional visual effects supervisors might use, Glick said; now they’re available to “everyday people” as well.

Many Hollywood studios see this technology as a way to streamline and cut costs when making movies. Greg Brockman, the president and co-founder of OpenAI, which makes ChatGPT, has argued that AI will improve cinema, allowing people to have immersive, interactive experiences with art. He gave the example of the finale of HBO’s “Game of Thrones.”

“Imagine if you could ask your AI to make a new ending that goes a different way,” Brockman said in March at the South by Southwest conference in Austin. “Maybe even put yourself in there as a main character or something, having interactive experiences.”

They thought loved ones were calling for help. It was an AI scam.

Nelson floated the idea of making a film using Dall-E images to OpenAI in September. The organization provided grants to support “Critterz,” but representatives of the company did not say how much they gave.

While Nelson created a bulk of the character images using Dall-E shortly after the tool’s launch early last year, he said he could not solely rely on artificial intelligence to create the movie.

AI cannot generate video with professional quality, Nelson said, so he would need professional animators to bring his images to life. He partnered with the production company Native Foreign.

Nelson also collaborated with his 21-year-old son, who programmed the 3D motion tool, called Unreal Engine, to animate the characters’ faces.

On Monday, the movie was released to the public, and OpenAI said it’s a template for how AI will democratize moviemaking. “Critterz is a vibrant example of how artists can use AI tools to unlock ideas that were once out of reach because of budget, time or resources,” Natalie Summers, a communications professional at OpenAI, said in a statement.

But that democratization does not come without risks to moviemaking and adjacent industries such as video games, according to media insiders and news reports. In China, artificial intelligence is already taking jobs away from video game illustrators in favor of software that can animate images in seconds, a report in Rest of World shows.

Nelson said it’s likely that AI will replace some jobs in Hollywood while also potentially creating more. He pointed to the entrance of computer-editing software, and how that replaced more manual movie-editing jobs and processes.

“There are some jobs that might just go away entirely,” he said. “There might be some pain, but through it all, I think there’s just going to be more opportunities.”

Media and legal experts also said the use of AI in filmmaking raises several concerns — and the law is still unclear.

Actors are taking a stand on the issue. Reeves told Wired Magazine in February that he’s had provisions in his contract to prevent the digital manipulation of performances since the late 1990s. While he sees the benefits AI can bring, he views it as more of a threat to Hollywood’s creators than a boon.

In the past, there have been hiccups using AI that still shape how people think about it today. Notably, Glick said, there was an incident two years ago in which a documentary about the late chef and TV personality Anthony Bourdain came under scrutiny because filmmakers re-created his voice and it was unclear whether they got permission to do so.

The current Writers Guild negotiations offer an early example of how artists and actors might try to protect themselves, Glick and Meyer noted. The union has stated that work generated by artificial intelligence may not be considered “source” or “literary” material, two key provisions that partially determine how credit is given to script writers and how they are paid, according to media experts.

“Companies can’t use AI to undermine writers’ working standards including compensation, residuals, separated rights and credits,” the guild said in a March 22 statement on Twitter.

But deeper legal issues remain, according to Meyer. Software such as ChatGPT, Dall-E and its successor, Dall-E 2, creates high-quality work because it analyzes patterns in massive data sets that contain intellectual property, such as images other artists have made or books and movies people have written.

How much credit is owed to people whose work is used in AI training data for is still an open question that copyright law has not answered, according to Meyer.

Some hints are coming from the government, he said, pointing to a mid-March directive from the U.S. Copyright Office stating that work created by AI without human intervention or involvement cannot be copyrighted. But there are several ongoing court cases that will likely offer stronger guidance, he added.

“There will always be early adopters,” Meyer said. “But I think [Hollywood] will take things fairly slowly until they have more information from the courts.”