How AI Is Actually Used In Video Production Today

There's a gap between what AI can do in a demo and what it can do in real production. We see the demos every day. Gorgeous renders. Mind-bending visuals. The promise that everything is about to change. But when you ask "what's actually shipping for real brands?"... the room gets quieter.

Jonny Glines has answers.

Jonny is a creative director who helps run dxp.pro, and as a designer he’s worked with A24, Arcade Fire, Steve Aoki, and Bruno Mars. He recently joined us on Building for Others to talk about how AI fits into professional creative work today, including a project with Mustang that shows exactly where the line is.

The Pattern That Works

Across immersive and concert-scale video work, Jonny sees AI succeeding in three specific ways:

1. Prototyping before the expensive part starts. In experiential work, the pitch phase is everything. AI helps you rapidly explore style, composition, and atmosphere so stakeholders can say yes or no before you've spent weeks building.
   
   
2. Technical utility in large-format pipelines. 4K and 8K deliverables punish render times. AI upscaling (used carefully) can be a pragmatic choice: render a cleaner base at manageable resolution, then upscale for final delivery when timelines and budgets can't support full-res brute force.
   
   
3. Selective production use that doesn't become the headline. Yes, AI can generate imagery that ends up on screen. But in brand-facing work, the question isn't can it. The question is where does it belong.
   
   

Mustang: A Case Study in Getting It Right

One project Jonny discussed was an immersive brand experience in a warehouse in LA's Arts District, telling the story the Ford Mustang across four decades. The 60s. The 70s. The 80s. The 90s, with projections on every wall and audio narration guiding visitors through the history.

To make each era feel real, you need people: dancers, bystanders, characters dressed in period-accurate costumes, moving through the scenes.

"To do a green screen shoot of hundreds of different people dressed in different costumes of different decades would have just not really been realistic with this type of project," Jonny explained. "Both in timeline, but also they've already spent so much money on the tech."

The solution: Jonny's team used Midjourney and Runway to generate AI characters that could be composited into the scenes. Hundreds of them. Different eras. Different outfits. Dancing, walking, existing in the world of each decade.

But here's the part that matters: they didn't try to pass off AI as the main event. The car was real. The environments were built. The AI characters were a layer, filling out the world, making it feel alive.

"If the AI is happening kind of in the background, if it's not the main focus, then it seems to work well with these types of projects," Jonny said.

A Simple Trick That Improved Realism: Put “Green Screen” In Your Prompt

One discovery from the project: prompt AI to generate footage on a green screen.

"We prompted them to be in actual green screen studios, which I thought was very funny," Jonny said. "The AI needs an anchor that's very strong. If you tell the AI to actually be in the green screen, it's going to give you much more realistic results."

If you try to generate characters on a black background or alpha channel and then key them out, the results fall apart. But when you prompt the AI to imagine a real green screen studio, complete with lighting and environment, the output is dramatically better for compositing.

Where This Fits

This project wasn't AI-first. It was a traditional production pipeline: After Effects, Cinema 4D, Unreal Engine, Topaz for upscaling. AI was one tool in the stack, used where it made sense.

This is the pattern we're seeing with brands that are actually shipping AI-assisted creative work. AI is used for rapid prototyping and handling the parts that would otherwise be impossible or impractical, while traditional craft handles everything else.

"This is what people were very interested in hearing about at the hackathon," Jonny said, referring to a recent AI filmmaking event where he served as a judge. "It shows an example of what brands are actually using for campaigns and where's the threshold of their comfort."

That threshold? Don't provoke the "icky feeling." AI in creative production isn't theoretical. It's happening. But it's happening in a specific way: as infrastructure, not spectacle.