Still experimenting: Building my digital fabric library

Yesterday, I posted about the system I’m using to prototype "Quiet Luxury" looks. Since then, I’ve been busy in the "AI studio" putting that system to work.

I wanted to see if I could get the AI to actually understand the difference between fabrics, not just the colors, but the physics. I spent the last day generating a digital "swatch-book" of 22 different materials, everything from heavy, liquid satin to rough, structured denim.

The main challenge: Most AI generators love to make everything look "perfectly smooth." To fix that, I stopped trying to describe the garment and started describing the behavior. I focused on how light should hit a fuzzy chenille weave versus a glossy silk.

Here is the prompt template I’ve been using:

"High-detail, editorial product macro shot of a stylized, matte-white mannequin with a featureless head. [Insert Fabric Type] fabric. Physics: [Describe drape, weight, and surface]. Lighting: Directional soft-box from front-left to highlight [texture/gloss]. 8k resolution, ultra-realistic, clean, professional."

It’s not just about making cool images. By keeping the lighting and the mannequin consistent, I’ve turned this into a real workflow. Now, I can swap fabrics in seconds to see how a design looks in different materials, which is way faster than waiting for physical samples.

The coolest part was seeing how the AI handles the "tension points" where the fabric pulls over the mannequin's knee or thigh. Those tiny creases are what make it look like a real garment.

If you’re experimenting with AI fashion design, try focusing on the "physics" of your material first. It’s been a total game-changer for my workflow.

I’m still adding to my library.