A controversy was brewing around “Young Washington.” Twitter users were sharing screenshots of the credits, which confirmed that the Jon Erwin-directed film included visual effects that utilized artificial intelligence. Others offered up freeze frames of shots they suspected were artificially generated. Then came the exclusive report in Variety: “Young Washington” contains about 100 shots augmented with generative AI, including some costumes and cannon fire.

So, how would Erwin respond to these accusations?

Mostly, with surprise, as Erwin (“I Can Only Imagine,” “I Still Believe,” “House of David“) is as vocal about using AI as any director currently working in Hollywood.

Before “Young Washington,” Erwin squeezed some AI into “House of David,” and this year’s “The Old Stories: Moses” had VFX that were largely AI. The filmmaker spoke openly about both projects at the recent AI on the Lot convention. His new studio, Innovative Dreams, has a massive volume stage on the same Manhattan Beach lot that also houses James Cameron, and they’re using AI to get something closer to real-time visual effects.

If you’re wondering if any controversy impacted the box office of “Young Washington,” the film made $20.8 million in its opening weekend from a $20 million budget, beat out “Supergirl,” is Erwin’s biggest opening weekend ever, and has already announced a sequel.

Erwin is not hiding anything when it comes to his embrace of AI. He believes it’s important to be transparent and lead a real conversation about how AI can and should be used in Hollywood. But perhaps more than anyone we’ve spoken with about AI, he’s an adult in the room speaking practically about what it can and can’t do, how it should be regulated, and how the rhetoric coming from so many in the tech space has been wrong.

“I think we can function at a level of cost that’s inexpensive enough and fast enough to create no-brainer decisions for studios. Within that context, a lot of things can be preserved,” Erwin told IndieWire from his studio space. “There are other people in this space that are almost celebrating how much they can replace. Why? Why would you create a race to the bottom to create inferior work? That makes no sense to me. Why is that a business practice at all? Why not say, how do we preserve? Why not have the same attitude that Steven Spielberg had when he looked at CG dinosaurs as opposed to stop motion and said, ‘Let’s do it that way, that’s so cool.’ That helps the industry move forward.”

Erwin feels strongly that AI can actually save jobs and bring production back to the United States, not replace it. With “Young Washington” and “Moses,” he now has actual test cases to prove it. “Moses” was made with the idea that they could film an entire series on a soundstage, while “Young Washington” was made almost entirely on location in Ireland.

Technology was applied to both in different ways, Erwin explained, with the volume stage backdrop on “Moses” allowing the project to transport its actors to locations they would otherwise not be able to go to, and with AI touch-ups allowing “Young Washington” to capture shots it would not have safely been able to otherwise.

Each production employed hundreds of crew and an ensemble of real actors, and each has the level of craft audiences have come to expect from Erwin. Each was also made copyrightable, ethically, and collaboratively between Erwin and his team. Each still required original, on-location photography that could be fed into the AI generative inputs, as well as with the AI tools built on top of traditional VFX software like Unreal Engine and Nuke.

Erwin is well-aware of the fear, skepticism, and vitriol that surrounds AI. Even if AI can do something, it doesn’t mean it should. Like any tech, there should be rules around it, and Erwin with his polite, Southern charm and light sense of humor might be as poised as anyone to guide where this conversation goes next.

“The reason why I really think these tools are going to last is because it creates a moment where creativity can happen as it should be,” Erwin said. “It creates a place for intuitive creation and collaboration. The filmmaking process, if you really think about the process, you really start to ask some questions of why in the hell have we done it this way for so long?”

Synthesizing — Not Generating — AI

It’s useful to understand just what on “Young Washington” used AI and how Innovative Dreams incorporates it into their workflow. Because as Erwin explained, they’re not really “generating” anything from scratch. Erwin generally uses the phrase “synthesizing,” in which the AI augments assets that they can prove they own and that have already been shot in camera, be it a piece of wardrobe, a location, or the actors themselves.

In some instances, AI was used on “Young Washington” to make it look like a cannon was actually firing during a battle scene, actively lurching back from the force, rather than just showing the gunfire burst. More crucially, it was used during a sequence set along an icy river while George Washington is traversing the uncharted Ohio Territory wilderness. They couldn’t put the actors into actual frigid water, so instead they filmed the real icy river and then separately filmed the actors inside a 50-foot hole or trench.

“You actually get to a much higher level of photorealism if you take real things that you can prove you own, the same way you own them like a chain of title to a script, like here’s my location agreement, here’s my actor’s consent, here’s my wardrobe. And when you use real things and you use these tools to synthesize those real things, you get to far superior results,” Erwin said.

The key difference on “Young Washington” is Erwin is using AI to create the visual effects in the moment, in real-time alongside that water pit with his actors, not months after the fact in post-production. Seeing the visual effects on his monitors can inform what he actually films. It’s a process they’re able to recreate on a sound stage in Manhattan Beach or out in the open air in Ireland.

Erwin sees this as a way to get his VFX department heads involved in the production process, not just isolated during post, and ensure the visuals being created match what’s being shot and how it’s being lit.

But even on “Moses,” teams are actually filming in Greece and Spain to capture the real locations that can then be projected onto the volume stage during filming. His crew took volumetric scanning cameras, a process with the unusual name “gaussian splatting,” to create the 3-D renders that could be seen, mapped, and projected onto the massive screen behind the actors.

But unlike on “The Mandalorian” that also uses volume stages, Erwin’s team can generate a new image in moments and do so for far less than hundreds of thousands of dollars. Erwin says it’s strangely “very relational,” and far different from his early career work as a 3-D animator and the tools he used then.

“One of the reasons I stopped is they were so difficult to control. They were so difficult to interface with. When you talk about keyframe animation in 3-D, you’re doing this by clicking a mouse and you’re trying to grab the hand of a character, and it’s so hard,” Erwin said. “The fact that you can actually interface with these programs, through language, talking the way you would talk to any department head on set, is magical.”

Innovative Dreams places its prompting computers on wheels to move from set to the editing bay and bring his AI artists generating right alongside the crew on set. But it also all works because Erwin is combining these AI tools on top of other more traditional software. A singular model like Sora alone might not have the right resolution or be able to create a whole movie, but using a combination of foundational models and other prosumer tools has for Erwin changed the game.

“People were only trying to use one tool. We realized we could stack them together in creative ways, and the stack was where the quality was,” he said. “We actually realized that this didn’t replace anything, it augmented what we were already doing.”

Empowering Professionals, Not Prompters

Erwin said something that’s rarely talked about in the context of AI and filmmaking is the cycle time between iterations. “How quickly can you get to the next version of the idea visually?” The longer it takes, you lose connection to the idea, and so much time is wasted trying to bridge the gap between a filmmaker’s vision and the visual manifestation of that idea.

Innovative Dreams’ hope is to take what Erwin describes as the bifurcated, slow, and linear style of filmmaking and turn that process on its head. Instead of going through prep, principal photography, and editorial and realizing you don’t have what you need, Erwin said AI makes “non-linear filmmaking” possible, allowing you to film, generate, and edit, all in the same day and time, and not necessarily in that order, something he believes is “creativity as it should be.”

“It doesn’t matter that it’s cheaper and faster. You can’t force a great creative to use a tool just because it’s cheaper and faster. The tool has to be better,” Erwin said. “Digital cameras had to at least get at a certain equivalent level of film. It didn’t matter if it was just cheaper than film. Now digital cameras do things that film will not do.”

Innovative Dreams’ vision of AI is for now only a dream because it requires the buy-in of the whole industry. Erwin supports a mandate from both the studios and unions to help re-train professionals on AI tools, and doing so he feels may be one of the best and only ways to bring jobs back to Los Angeles. He’s found through his own experience that training veteran film professionals on how to use AI tools is far more effective than teaching solo hobbyists making AI videos on their phones how to collaborate on a film set. He wants to further empower the people who have been working with him forever.

It also requires a conversation about the ethical application of AI tools, such as curbing things that would modify a performance or infringe on copyright. For instance, he draws the line at synthetic performers, not just for ethical reasons, but for practical ones.

He first believes the arms race has created a “false economy” within the AI community today, and the cost of doing an individual prompt will escalate over time. So while it might be way more cost effective to do explosions, chase sequences, or wide shots with AI, prompting an entire actor’s performance for a feature film is a different story. He also knows how special it is to see a real person perform words that he’s written, and that’s one thing that can’t be replaced.

“Any mindset of how many people can we replace is immoral, and it’s the wrong way to think,” Erwin said. “What we should think is, could we use these tools to combat the primary reason jobs have been lost, which is the escalating cost of production and the sheer time it takes to create?”

Erwin said the moment in time we’re in right now with AI is akin to when George Lucas was working on “Star Wars Episode II” in 2002. Lucas hosted a summit of filmmakers to talk about what does a digital pipeline look like in both the capture and distribution of film and how to take the industry digital. Filmmakers are beyond the shock of seeing AI explode so quickly into our lives, and we’re now debating how these tools should be used and who should be calling the shots.

“We’re the people that have dedicated our lives to this craft. We should not sit on the sideline. Let’s roll up our sleeves. It’s a lot easier to criticize something than it is to get curious and to start trying it,” Erwin said. “I think the best path to that is transparency and to create a conversation about it.”