One image, one prompt: a free Claude skill breaks a hummingbird into 300 parts

A photo of a hummingbird, one prompt, one free Claude skill. Out comes a 3D exploded-view animation, with 300 small components hidden beneath the shell panels and 7 glass blades on each wing.

Nyxera's demo on X opens with “F*CKING INSANE.” What surprised him wasn't the precision—it was that nobody is charging for this skill.

For this specific model, it was built with three.js. At 1.7 seconds, the wings unlock at their roots and flip over in midair, while the entire shot is slowed to 0.4x speed. The eye unscrews in 4 rings, and a 10-blade aperture opens from behind. The beak sheath pulls back, and a 1.8 mm probe extends 24 mm past the beak tip. After that, a new component opens every 2 seconds; the exploded parts rotate in the opposite direction, and it finally rises to a top-down view to finish.

The most easily overlooked sentence in the whole thread is that Opus first compares its rendered still frame with the original image, and only starts animating once the two line up. In a reply, user Foxik said he had skipped this step before and paid the price. That reply carries more information than the main post.

The skill and prompt come from an article by Elvar, which includes the skill, the prompt, and the simulator shown in the demo: Elvar's article. Nyxera reposted Elvar's post, and people in the comments said they wanted to install it, star it, and try it out.

The same skill has also been used to break down other things. A quoted tweet includes a technical schematic animation of an S&W Model 686 revolver, labeling trigger travel, hammer angle, and single-shot time. Bionic Lion commented below, saying it looks like DeepSeek could learn a thing or two.

There are two questions the main post didn't address. Lateascent asked whether a free Claude account can run this skill or whether you have to pay. Edward Jr asked how much of the effect comes from the skill and how much from the prompt. The first question is about whether you can get started; the second is about whether this is reproducible.

What's noteworthy is the workflow, not the generation. Most people see a demo like this and their first reaction is to ask for the prompt. But in this case, aligning a still frame with the original image before moving on is a hard constraint. It puts the judgment of “does it look alike” before the animation; if you wait until it's finished to redo it, it gets expensive. That may be where the valuable part of the free skill lies: it prescribes the order.

As for the fact that “nobody is charging” itself becoming a selling point—that's also pretty interesting. The tool is open source, the skill is free, and the result looks good, so the chain of spread starts turning on its own. Charging for it would only interrupt it.