Use case: Catalog magic with AI Studio

How a large shoe retailer uplifted their product catalog ads with Smartly’s AI Studio, and didn’t need weeks to do it

A large shoe retailer had a feed they used for their catalog ads that contained basic product images. As part of a brand refresh, they wanted to recreate their product imagery, and challenge themselves do something totally new with the assets they had on hand.

The starting point was a simple feed source file with product images, colors, style names, descriptions, and prices. Initially the company thought that changing the product images would need a lengthy, expensive photoshoot, especially if they wanted anything other than a basic studio background.

How wrong they were! With Smartly, they would soon have their entire product catalog enriched with new images, showing their products both with individual studio backgrounds and in real-world scenes!

Tailored background

After taking Smartly’s AI Studio into use and experimenting with it, the shoe retailer’s marketing team realised they could very easily remove the existing background from a large number of photos very fast. This took away some of the pain, but the next question was, what would they put in the background instead? Maybe a color? Maybe some text, too?

What was more, they also didn’t want to spend days picking colors and copy-pasting pictures of shoes.

So they did something more innovative…

Getting color codes

The catalog of shoes already included information about the color of each shoe style. They were expressed as color names - words like “red”, “purple”, “tan”, and “cool grey” - and the team needed the corresponding codes to proceed.

First they used Text Generation for the slightly unorthodox purpose of generating color codes from color names. They crafted a prompt that would output only the corresponding color codes.

Here is their prompt:

Please give me a color code that matches the color in {{color}}. I need only the color code, nothing else, in the format # and the code so that I can connect it to my image editor. I need the color not to be too bright, but still well saturated, because I want to use it as a background of a product and I want to keep the focus on the product.

Then, to get the text elements to stand out, they asked AI to match each color to either black or white text, depending on which provided the greatest contrast:

I need to add some text on top of a rectangle with {{ai_color}} color. Give me a white or a black color code  based on which you think is easier to read a text placed over the rectangle with {{ai_color}} color. Give me only the color code of your choice, nothing else.

Once they ran these through Generation on Feeds one after the other, they had a feed file enriched with two color codes per item.

Generating the backgrounds

The next step was to have Scene Generation remove the existing backgrounds. This was a quick task.

Then they went over to the Smartly Template Editor and set up the background to use the color code now available in the feed source. This gave each shoe a matching background automatically. To give the ads more energy they posed the shoe image at a 45-degree angle as well.

The text is also a dynamic element, created in the Template Editor. To get its color to match the background, they referenced the text color code field that AI Studio had added in the feed source file.

By the way, the drop shadow is a dynamic element, too. That’s how it changes to match the shape of the shoe.

Where did the tagline come from?

Smartly AI Studio, of course, with the following prompt used on the shoe description:

Create a short tagline. It needs to be short as my textbox is really small. Max 4 words.

Background scenery that matches

But they weren’t done yet! While the studio-like images were very nice now, they wanted their ads to really pop. What if they were to let consumers see their shoes in a lifelike setting, in the actual environment where they would be used?

Getting suitable descriptions

The product catalog also has a description of the product features. First they asked Text Generation to write a prompt for image generation based on that description:

Create a prompt for AI to create a background image. Make it less than 200 characters. I only need the description of a possible background for shoes. It's very important that there are no people, I want the focus only on the product. Do not mention the name of the product, call it “a shoe”.

And the next step was, of course, to generate the images.

And now, their product catalog can also feature the shoes in their natural environment!

Ready for ads

But catalog ads come in three different formats. To accommodate multiple channels and formats, they used Generative Expand to create a bunch of images in other aspect ratios.

And the finale - video magic!

When these results were presented to leadership, they came back with an ask (of course they did!) for videos. After a brief moment of panic the team got in touch with Smartly, who pointed out that they could create videos by animating static assets in the Template Editor. And so they did.

Needless to say, leadership was more than pleased with the final result, and the very short time in which dozens of smashing product videos were produced!

P.S.

If you try out the prompts in this article for your own images, bear in mind that Smartly continually develops the AI Studio. The same prompt probably won’t give you identical results. Keep experimenting with your prompts, and if necessary, use another AI LLM to help!