POV: AI is at its best when combined with traditional processes

A frame-by-frame animator, you’d imagine, would curse the ground that generative AI stomps upon. So it was surprising to hear London-based director Alice Bloomfield tell a crowd at Nicer Tuesdays how she used AI on one of her favourite animation projects: a music video screened at London’s immersive Outernet space, drawn by hand in the traditional frame-by-frame style.

“[The music video] was all done. I had spent three months inside, working away,” Alice recollected to the audience. “We had one last meeting with the Outernet team. I remember them saying: ‘It’s all looking good, but we’ve had a few last-minute checks and I think we’re gonna need a higher frame rate. You’ve been using 12 frames a second, but I think we’re gonna need at least 24 frames per second.’”

At this point, Alice was “freaking out”: “I’d just spent three months in my room, working, and now you want twice as many frames?” As an independent animator working not from a large studio but from her bedroom, the idea of locking herself away to effectively double the workload, just to meet the resolution requirements of a large technical screen, seemed fruitless indeed. Interestingly, the solution to achieve the high-fidelity demand came by using another form of technical trickery. By working with an AI company and doing some light clean-up by hand, Alice was able to generate a new frame in between every existing frame, smoothing the viewing experience for the Outernet audience.

Alice’s story comes from 2023 – when the shockwaves of AI were mere ripples. While it was a novel solution for Alice and the Outernet team, today, AI’s presence in film is widespread. As Jim Geduldick, a VFX supervisor and cinematographer working on major Hollywood pipelines like Disney’s live-action Pinocchio, recently told The Guardian: “Everybody’s using it. Everybody’s playing with it.”

In this landscape, the potential negatives of AI are growing. The compulsive use of AI is threatening roles and careers within the film industry, which is a space already defined by mistreatment and unrest. What’s more, in film and visual campaigns, the word AI conjures one specific connotation: slop – otherwise known as, mass-produced, untasteful, lazy – standing in direct opposition to tradition or craft.

This is why it is hard to not be hopeful – and maybe, a little conflicted – when you see the work of Jeremy Higgins, a director and animator whose work can be described as both soulful, and AI-driven. His process is what most would regard as “traditional”, often working with hand-drawn 2D animation and frame-by-frame processes. But he’s also known for blending these traditional pipelines with other technology. In 2024, working on a Khruangbin music video, Jeremy and animator Jenny Lucia Mascia drew 3000 individual frames using pastels. While these frames were technically hand-drawn, Jenny and Jeremy drew them on top of an entirely CGI-generated 3D animation, leading to a unique blend of technical processes and old-school artistry.

He’s trying out similar things now with AI. “I try to combine traditional animation with AI in as many ways as I can come up with,” he says. “It started with stop motion of AI images, then I started doing character animations over them because I saw interesting potential to create unique-feeling settings. I’ve played with printing the frames and drawing over them to add textures. Really whatever I can do to add more of the human touch and just generally make something that feels more like me, rather than it feeling like pure AI.”

When Jeremy uses AI for animation, it often involves a certain amount of “stepping in” to tease out the soul. In Migration, a short film Jeremy made in 2024, Jeremy and his team drew characters with traditional cel animation, and used AI to generate thousands of background scenes, before coming back in to animate contact shadows and interactions between character and background.

For Jeremy, it’s through the combination of home-grown visuals and generative technology that the best results with AI are produced. “I will say,” Jeremy says, “if you AI-generate a whole ad with nothing additional brought into it, I think it’s going to show. If you want to use AI and make something truly exceptional that you can be proud of, people should treat it like a new step in the production pipeline. Yes, AI can make good images easily. So now it’s like okay, how can we use that to push creativity further?”

Jeremy raises a vital point: that soulful or “exceptional” AI-led creative work is best achieved when the human hand is present in the final output. But, is there also soul to be found in AI work that goes to great, experimental lengths to be bespoke, tailored and unique? Where the process of working with AI becomes a creative production in and of itself? I think so, and so does Paris-based creative studio Unveil.

Generative AI is typically thought of as exactly that: generated, as if from thin air. But Unveil describes the process of building a recent dataset for a Heliot Emil x Puma as a “full production on its own”. Don’t get them wrong, this campaign film – which visualises the experience of a “runner’s high” through images of floating athletes – is fully AI-generated. “None of the final campaign content was shot or filmed,” says co-founder Alexis Foucault.

But, it’s also highly bespoke. The team created two separate datasets for the campaign – “one to train the model that would generate the images, and another for image retouching,” says co-founder Tom-Jacques Perret. For the first, Unveil shot hundreds of photos of the campaign models “levitating”, just to teach the AI how to understand the complexity of movement the campaign was aiming for. “For the second, we used detailed packshots of the products photographed from every angle, with precise fabric details, so we could regenerate them perfectly later,” said Alexis.

In Unveil’s studio, generative AI is a service offered alongside many others – CGI, digital design, graphic design, content creation. The blend of traditional production and generative AI is something they are actively experimenting with. “The easiest overlap is when you shoot specifically to create a high-quality dataset for a project, which you then use to train a custom model and generate the final assets,” says Tom-Jacques. Using generative AI as a tool provides Unveil with one key ingredient in its process: randomness. “The biggest difference is that AI often leads us to discover new compositions or aesthetics we didn’t imagine at the start; it’s an organic creative tool, almost like having another person in the room,” says Tom-Jacques.

“Soul” is not easily quantifiable. Is CGI “soulful”? Most would argue, yes – it involves the hard-earned labour of a creative that has cut their teeth within a distinct visual discipline; it’s as soulful as any other creative pursuit. In its current form at least, it’s hard to draw an equivalent between CGI and AI. AI is a technology leading to widespread corner-cutting, as well as the culling of jobs from the creative and production pipelines. As Alexis says, “[AI] won’t wipe [advertising campaigns] out, but it will shake the industry hard. We’ve had conversations with producer friends who are seriously worried and they’re not wrong. We believe brands will produce more campaigns each year, reserving traditional shoots for key moments.”

The work of people like Jeremy, Alice and Unveil does not quell the industry-wide threats ushered in by the AI era. However, if VFX supervisor Jim Geduldick is right, and AI is already being tinkered with Hollywood-wide, we’d hope that a reality is possible where human hands can successfully dance with a dataset. In the case of AI assets, it will be those that go against the expectations of AI – fast, cheap and easy – and that deliberately impose limitations on their usage, that will maintain that coveted, elusive substance: soul.