3D Modelling with AI - A Scale Modellers Guide

AI‑assisted figure creation has really surged in popularity lately, and I’ve been experimenting with a range of tools to support both my current builds and the projects I’ve got lined up next. The technology is moving quickly, and it’s been fascinating to see how different AI workflows can help generate custom figures that would have taken far longer to sculpt or kitbash by hand.

As I’ve worked through some of these tools, it struck me that a clear, step‑by‑step walkthrough could make the whole process far more approachable for other modellers who are curious but unsure where to begin. That’s what this guide aims to do: break the workflow down into simple, repeatable stages so anyone can follow along and start experimenting.

I’m not presenting myself as an expert—just a modeller sharing what’s been working so far, the pitfalls I’ve run into, and the small wins that make the process enjoyable. If this guide helps even a few of you feel confident enough to try AI‑generated figures for your own projects, then it’s done its job.

I’ve put together the following slides to walk you through a simple, real‑world example of how I took a single 2D image and, with the help of a couple of AI modelling tools, turned it into a brand‑new 3D STL file that can be printed at home. The goal isn’t to show the only way to do it, but to demonstrate a clear, repeatable workflow that anyone can follow.

To make the process easier to try for yourself, I’ve included all of the prompts I used as selectable text beneath each slide. You can copy them directly, tweak them, or use them as a starting point for your own prompt variations. The idea is to give you a practical reference you can follow step by step without needing to reinvent the wheel.

Things to Consider

Oh, it's also worth mentioning that AI isn’t perfect, and it certainly isn’t consistent—and that’s completely normal. Even if you use the exact same prompts I’ve shown, word‑for‑word, on your own image, you may still end up with different results. That’s not a mistake or something you’re doing wrong; it’s simply how these systems work.

Each time an AI processes a prompt, it “thinks” about the request in a slightly different way, drawing on probabilities rather than fixed rules. That means the output can shift from run to run, even when nothing else changes. Sometimes those variations are tiny, sometimes they’re dramatic—but they’re all part of the nature of generative AI.

The key is not to expect identical results every time, but to treat the process as iterative. If the output isn’t quite what you want, a small tweak to your wording—or even just running the same prompt again—can produce something much closer to your goal. Embracing that unpredictability makes the whole workflow smoother and a lot more enjoyable.