Kaufland's "Ice Ice Baby": A Case for Hybrid AI Production

Kaufland has been here before.

In early 2025, the retailer released a fully AI-generated spot for Gönrgy. It made headlines because it was created with AI. That was the point — and also the ceiling. When the story is simply that a machine made the ad, the machine becomes the star rather than the brand. One year later, Kaufland is back with another AI campaign.

Except this time, almost everything viewers emotionally connect with is real. The new Kaufland AI campaign, called “Ice Ice Baby,” was created by EASYdoesit around one gloriously stupid-brilliant idea: eat Kaufland’s ice cream and transform into Vanilla Ice. The hair. The tracksuit. The moves. A grandmother, a father, a child — everyone gets their Vanilla Ice moment. How the spot was made says more about the future of advertising than most AI keynotes ever could.

Inside the Ice Ice Baby Kaufland production

The production started with 12 real actors. Stylist Marianka Benesch created three physical Kaufland signature tracksuits. Every other suit was generated with AI and fitted onto the actors during post-production. Before the shoot, the team also used AI to explore and build 23 different worlds that no location van could ever reach.

Those environments were first developed as still images, then retouched and translated into motion. On set, multiple units worked in parallel across video, photography, physical props, and greenscreen production. The generated environments were displayed live on the VTR, allowing the gaffer to match the real lighting to locations that only existed as digital files. Read that again. The AI was not replacing the crew. It was displayed on the monitor, giving the crew something to light against.

The playbook: Shoot the humans, generate the worlds

The 2025 AI advertising playbook was straightforward: generate everything, reduce the production budget, and benefit from the novelty. The traditional production playbook takes the opposite approach: travel to every location, build every set, and hope the weather cooperates. The hybrid AI production playbook sits between the two. Shoot the humans. Generate the worlds. Composite the truth. That distinction matters.

Audiences have become increasingly good at recognising fully generated advertising. They notice the rubbery faces, the physics that almost work, and the emotions that almost land. What is far harder to detect is a real performer, filmed by a real director of photography, biting into real ice cream inside a world that has been generated, retouched, animated, composited, and graded until it earns its place in the frame.

As the team explains in its behind-the-scenes film, artificial intelligence makes the production possible. But so does the intelligence of real people who have spent years learning what a shot needs behind the camera. The people stay real where people matter. The worlds become synthetic where physical worlds would simply be expensive.

AI moves behind the camera

There is another detail in this production worth pausing on: the role of the AI Creative Director. During post-production, director David Aufdembrinke worked alongside AI creative director Caspar Jade. One owned the film. The other owned the machine.

That pairing offers a more realistic picture of where the industry is heading. The director is not replaced. Instead, the director gains a counterpart who understands the models, their possibilities, and their limitations.

The role is not only about knowing how to generate an image. It requires understanding how the generated material fits into the direction, lighting, composition, performances, and rhythm of the final film.

The credit list quietly resolves another ongoing industry debate. Five AI artists — Florian Feldmann, Jonas Hofmann, Milad Mesof, Malte Milbrand, and Robert Schrödter — worked alongside compositors Bernd Macht, Martin Pedreira, and Julian Trunke, as well as retoucher Marius Schwiegk. AI artists and VFX compositors were part of the same production pipeline. Two crafts that are often presented as enemies worked together to make one film.

This is one of the most important aspects of the EASYdoesit production. AI was not treated as a separate department responsible for generating finished shots in isolation. It became part of a broader post-production process shaped by retouching, compositing, lighting, editing, and direction.

Hybrid production is still production

None of this means the process is easy. Hybrid pipelines can be messy. Matching the lighting of a studio performance with a generated environment requires craft, not magic.

A workflow involving 23 AI-generated worlds still needs someone who understands why world number 14 feels wrong — even when the image is technically impressive. Someone still has to recognise the problem, communicate it, and decide how to fix it.

That ability may become one of the most valuable production skills of the next decade: understanding what the technology can produce while also recognising when the result does not serve the film.

From an ad about AI to an ad made with AI

In 2026, Kaufland made an ad with AI, about ice cream. Together with EASYdoesit, producers Dimitri Hempel, Maxim and Alireza Davatgar, and a crew that clearly had fun doing it.

That is the whole shift. And it is why hybrid wins: the technology moved out of the headline and into the toolbox. It will belong to teams that know which parts should remain physical, which can become synthetic, and how to combine both without losing the emotion of the idea. That is why hybrid AI production wins.

Word to your mother.