🍎 Slicing a LEGO Apple in Half - VEO 3 Test

Eight seconds is not much time to prove anything, but it is enough to test whether an AI video model can pull off one very specific kind of satisfaction: the clean slice. This clip takes a custom-built LEGO apple, complete with a realistic core, tiny black seeds, and cream-colored flesh, and cuts it neatly in half to see whether a Veo 3 generation can deliver that oddly satisfying payoff people chase in slicing videos.

Why a LEGO apple

Slicing videos work because of a very specific promise: something clean on the outside opens up to reveal something detailed and unexpected on the inside. A LEGO apple is a good subject for that promise precisely because it is artificial. Nobody expects a plastic brick model to have seed placement, a defined core, and distinct interior coloring, so seeing all of that appear the instant the knife passes through is what makes the moment land. The build itself, before any AI was involved, had to get those details right: the seeds placed where an apple's seeds actually sit, the flesh colored to look edible rather than like undifferentiated plastic.

What the test was checking

Labeling the clip a "Veo 3 Test" signals the real point of the exercise. This is less about the apple and more about the model. A slicing motion is a hard thing for a generative video system to get right: the blade has to move smoothly, the two halves have to separate cleanly, and the newly revealed interior has to stay consistent with what a real cut cross-section should look like, rather than warping or smearing as the camera holds on it. Testing that on a small, self-contained subject like a single piece of fruit makes it easy to judge, in eight seconds, whether the motion and the reveal both worked.

The appeal of oddly satisfying content

The genre this clip sits in, oddly satisfying content, has its own quiet rules: no distractions, no dialogue, just a clean action executed well. That makes it a useful proving ground for a video model, because there is nowhere to hide a flaw. A shaky cut, a seed in the wrong place, or an interior that does not match the exterior would be obvious immediately. A clip like this either satisfies on first watch or it does not.

Key takeaways

• The clip is an eight-second test of Google's Veo 3 video generation model, built around a custom LEGO apple.
• The subject was chosen for the classic "satisfying slice" payoff: a clean cut revealing a detailed, realistic interior.
• Interior details, like seed placement and flesh coloring, were built in from the start rather than left generic.
• Short, single-action clips like this are an efficient way to judge whether a generative video model handles motion and consistency well.

Who this is for

This one is for anyone curious about what current AI video tools can and cannot render convincingly, and for LEGO builders who enjoy seeing a physical build idea tested against a generative model. It is a quick, low-stakes look at Humanitarians AI's ongoing experiments with creative AI tools, not a tutorial or a deep technical breakdown.