4 methods to generate fashion ideas with image AI

Published Categorized as Tutorial Tagged 2 Comments on 4 methods to generate fashion ideas with image AI

Stable Diffusion AI distills millions of images in the model. In layman’s term, it has seen a lot. If creativity is new arrangement of knowledge, it surely has potential to create.

In this post, we will use the image AI to generate new fashion clothings. The AI will show us something we have never seen, and we can use them as ideas for our own projects.


We will use Stable Diffusion AI and AUTOMATIC1111 GUI. See my quick start guide for setting up in Google’s cloud server.

Select a starting image

You will first need to get a fashion model. The best is to get a full-body image with a simple monotonic background. Below is what I am going to use in this tutorial.

Starting fashion model.
Original image

Create inpaint mask

In the Stable Diffusion GUI, go to img2img tab and select the inpaint tab.

Upload the starting image by dragging and dropping it to the inpaint image box.

Then create a mask covering the whole body like this:

Use crop and resize, and set the image size to aspect ratio similar to the original.

In the methods below, you may want to explore changing CFG to strike a balance between freedom and following the prompt.

Method 1: Describing the apparel items

Use terms describing what the model is wearing is an effective way of generating a specific clothing.

For example, images below are using a single-word prompt “sweater” for inpainting. (Euler a, 35 steps, denoising strength 0.9, CFG 7).

But you don’t need to stop there. You can throw a kitchen sink at Stable Diffusion and let it “figure out”. Below is a prompt that includes multiple fashion items.


colorful, wedding dress, Sweater, T-shirt, Sheath dress, Dress, Gym clothes, Bra, Tank top, Shorts, Hoodie, Jeans, Long coat, Uniform, Coat, Dress pants, Swimsuit, Long-sleeve, top, Skirt, Thong, Hat, Necklace, Ring, blue, yellow, green, black, Earrings,Perfume, Mittens, Sunglasses, Handbag, Bracelet, Purse, Watch, Umbrella, Scarf, Slip, High-heeled shoes, Flip-flops, Boots, Stockings, Jacket

The results are definitely interesting:

When using diverse apparel terms and relatively low CFG (7), we allow Stable Diffusion AI to pick and choose the combinations that makes most sense.

Method 2: Apparel types

We can use apparel categories as prompt. That avoids being too specific and can create an element of surprise.

Below are generated with prompt “Sun dress”.

Here’s a list of other apparel types, just to give you some idea!

  • Culotte dress
  • Babydoll dress
  • Wrap around dress
  • Kimono dress
  • Bubble dress
  • Tutu dress
  • Polo dress
  • Shirtdress
  • Sundress
  • Sari
  • Sack dress
  • Tunic dress
  • Sheath dress
  • Shift dress
  • Trumpet dress
  • Smoked/layered dress

Method 3: Emotional words

Somethings it is better to say how great it is without saying what it is. Let Stable Diffusion figure out the fashion.


elegant, stylish, stunning

(denoising 1, cfg 15)

Method 4: Occupation

Let’s include uniforms in our fashion brainstorming. The trick is to think about what kind of people wear uniform for their work.




Flight attendant

Some other ideas are

  • Firefigther
  • Pumber
  • Maid
  • Chef
  • Pilot
  • Lifeguard


We have covered using inpainting to create different fashion styles using the free and open-source Stable Diffusion AI. Describing the fashion items directly is a powerful method but not the only way.

Sometimes it makes sense to be less specific and let AI explore the space.

Hope you will find this tutorial useful.


By Andrew

Andrew is an experienced engineer with a specialization in Machine Learning and Artificial Intelligence. He is passionate about programming, art, photography, and education. He has a Ph.D. in engineering.


  1. This is interesting! How to do with prompt if I wanna have the girl wear oversize fashion? It was always fit size to me, was strengthen the prompt the only way? Like (((((((baggy_shirt))))))) ? Thx, appreciate any reply ?

Leave a comment

Your email address will not be published. Required fields are marked *