When McCann asked for a meditating cow to introduce their new no-sugar, no-fat, lactose-free milk, we knew it would be a fun challenge — and a good chance to blend generative AI with traditional 3D and compositing. At first, we even joked that this would be the easy kind of project… the “just press the AI-magic-button” kind.
Spoiler: it wasn’t.
Zero control – 8 sec max – useless
Full control – 20sec (sped up) – meet clients needs
One of the biggest hurdles with generative AI is control and consistency. Every iteration behaves differently, which makes continuity a real puzzle. To tackle this, we generated multiple AI clips of the cow—full-body shots, close-ups, even specific details like ear twitches—and upscaled everything for maximum flexibility. From there, we built 8K composite for the background and the cow itself, mixing and stitching footage to craft a unique 20-second performance. It required a lot of morphing, 2D tracking, puppet animation, and clever fixes to overcome the quirks of each AI-generated fragment.
To bring stability and proper cinematography to the piece, I recreated the scene in 3D with depth, mapped all the AI elements onto it, and animated a smooth, controllable camera move—solving the unpredictability issue while keeping the natural look of the AI footage. The final packs and table are fully 3D as well.
Everything was then brought back into compositing, layered and polished for extra magic.
A surprisingly technical project disguised as a very chill cow doing yoga — and one that required far more human craftsmanship, problem-solving, and manual work than anyone would ever guess from the final zen moment on screen.
