One image in, cocktail explainer video out: Opus 5.5 renders it all in HTML

Ror_Fly · x · 2026-09-24

A user fed Opus 5.5 a single image and the model produced a full HTML/JS motion graphic: a 30-second explainer-style animation showing a cocktail recipe from empty glass to finished drink, with ingredients and measurements appearing as they're poured.

The prompt's key points: let the model pick HTML or JS, explicitly require the ingredient list and measurements, and time the animation to the pour order. The author calls it "getting weird" — a showcase of frontier models understanding recipe structure from one image and outputting finished animation.

Original post →

More from Fun

Fun channel →