Today, I’ll be trying to answer a question that’s been top of mind for me. Can I use AI to create visuals for ANY wearable product? And I mean ANY, I'll be testing a few that are notoriously difficult to photograph effectively (even under normal conditions) for fashion brands, and we'll include some more unconventional apparel as well.

Introducing Product to Model

What is Product to Model?

Put simply, Product to Model is a workflow on FASHN's Studio that transforms product images into people wearing those products. You give it an image, and it creates a product image for you in any aspect ratio that you choose.

You can optionally add a prompt to the generation or use AI models from the FASHN platform to generate more specific images. I'll show you how it works below.

The Workflow We'll Use

The process is simply "upload, generate, review." - and repeat until satisfied. My goal with this experiment is to see if someone with zero photography skills (myself) can get professional results without all the upfront investment of money, time, and learning. Let's get straight to it.

Product to Model Walkthrough

I click into FASHN’s Product to Model page in the Studio.

I want all of my results to be 1:1 (Square) aspect ratio, so I select it from the dropdown.

The “Add Model” button lets you generate or select an AI model to put your product on.

We also have a text input that allows you to describe in natural language things that you want to see in the final outputs of the visuals we create - for example, “blonde hair” or “man in park”.

Putting It to the Test

To keep the rest of this experiment fair, I’m only allowed to hit “Run” once for each product that I’m testing.

Suit

We'll start with a suit on a mannequin. Suits are already complex and confuse previous generation AI models. Making this even harder, the suit is on a Mannequin which adds to the difficulty, as the AI model has to cleanly differentiate the suit from the Mannequin to turn it into a nice product shot.

The top right generation is the raw output of the Product to Model workflow. Bottom left is the output after we added "at office". Bottom right is generated using a consistent AI model from the FASHN platform as guidance for the product image.

We're off to a good start; we see that Product to Model does great on its first try.

Backpack

Next up, we'll try an accessory - specifically, a backpack.

The top right generation was the raw output of the Product to Model workflow. Bottom left was the output after we added "at gym". Bottom right was generated using a consistent AI model from the FASHN platform as guidance for the product image.

Great results again, the AI even thought to position the model to be facing away from the camera to show the bag off better. Nice!

Hat and Polo (2 Items)

Let's throw it a curve ball to see how it performs with multiple items in one.

Fanny Pack

Now, another type of bag - a fanny pack.

The composition of the photo is great, and it nailed the zippers designs, as well as the texture of the leather.

Baby Onesie

Baby photos typically perform poorly with previous generation AI models. Let's see if Product to Model can accurately create baby faces.

I'd say that Product to Model workflow nailed the baby face in this. What do you think?

Sari/Saree (4 Items)

Next, I'm going to be trying out ethnic clothing. We'll start with a Sari/Saree, a style of clothing that originates from the Indian subcontinent. Most AIs struggle with it due to the complexity of its draping and folds. Let's see how the Product to Model workflow performs when I give it a full set (multiple items).

Note that we threw an additional handicap at the model here by giving it a flat lay (and not even a great one at that), with different items of the Saree folded up and layering over each other.

The product shot here isn’t the best, with a bird's eye view of just enough information to understand how the shirt, pants, shoes, and pallu should look on a person. I'll let you be the judge for the result. How do you think it did? I think it did great, especially since the product photo provided was not the greatest.

Ranek û Çoxe (6 Items)

At this point I'm trying to really test the AI's limits and break it. Here, I tested Product to Model with a Ranek û Çoxe, a traditional men's outfit from the Caucuses. AI models struggle to render this attire accurately because its style has countless hyper-specific variations from town to town, a cultural nuance that is often missing from AI model training data. It's also a multi garment product, with a large accessory (sword). Let's see how it does.

This blew my mind - if it can get this right, I have a hard time believing that there’s anything that it can’t. All the initial skepticism that I had around the Product to Model workflows capabilities have evaporated.

How You Can Use This Yourself

When I started this experiment, I wanted to see if someone with zero photography skills could create professional visuals for any product. I tried to find the Product to Model workflows limits by testing it with the most difficult items I could think of. After throwing everything from a suit on a mannequin to complex, multi-piece cultural attire at it, the results speak for themselves.

You can finally create the visuals you see in your head without insane budgets - the power to create is now more accessible than ever.

Access Product to Model right now in the FASHN App's Studio and start creating.

I can't wait to see what you make!