Opus 5.5 builds an immersive Italian 3D scene from one prompt
Using another project as reference and a single prompt with no iteration, Opus 5.5 generated a 3D scene with lighting and transitions in 419K tokens, which the author says already feels like a finished interactive experience.
419,000 tokens. One prompt. No refinements. And Opus 5.5 built this. I gave it Meng’s project as a reference and one detailed prompt to reimagine the whole experience in an immersive Italian setting. This was the first result, with no follow-up prompts and no extra iteration. The 3D, atmosphere, interactions, lighting, and transitions already feel way more complete than I expected from a first pass. It doesn’t really feel like a rough AI-generated concept anymore, it already feels like an actual interactive experience. And yes, 419,000 tokens sounds like a lot, but for something this complex, I was honestly expecting even more, especially compared to what I probably would have used with Fable 5.1. About 3 hours of active work overall, and I still haven’t started refining the smaller details. If this is only the first pass, I’m really curious to see how far I can push it.