[Image Generation AI] Creating Fashion Photos for Specific Purposes with Midjourney!

[Image Generation AI] Creating Fashion Photos for Specific Purposes with Midjourney!

Hello. I'm Sasaki, an intern.
Today, I'll be exploring whether we can reproduce fashion photos commonly used in the fashion industry using the image-generative AI, Midjourney.

Settings in Midjourney
Since this is an exploration, I've set it to "Remix Mode" to easily modify the prompt little by little. With the brand concept in mind, I've also set the stylize value to "high".

For more details on "Remix Mode", please refer to the Midjourney official website.

 

Fashion Photos Under Review

Among those predominantly used in the fashion industry, I've decided to review the following two:

Lookbook: A lookbook is similar to a catalog where one can view the entire collection, representing the brand's theme and vision for that season. It captures models wearing the items, showcasing how they look and can be paired. Typically, the styling reflects the brand's perspective, serving as a photographic compilation for viewers to understand and feel the brand's world and story.

For E-commerce Sites: Photos for e-commerce prioritize highlighting details and texture to facilitate purchases. It's essential to capture the color, texture, and intricate design details—elements crucial for consumers when choosing a product. Usually, items are photographed against a plain background, sometimes alongside shots of models wearing the items.

Assuming these photos were taken for the "neomode line" of "accelerando.Ai", I aim to generate images in their style.

About accelerando.Ai:

"accelerando.Ai" is a novel fashion brand rolled out by open fashion. Its core concept is "intersection."

 

↑ Image representing the Neomode of accelerando.Ai


I'll generate photos for the "Lookbook" and "E-commerce" that align with this brand's concept.
For this exercise, I've set the prompt to:

"one model ,Unique details, virtual fashion,Near future,Curvilinear details ,simple background"

 

After a few trials, some fabulous outfits were generated.

 

Making it More Like a Lookbook

 

 

 

I'll proceed to generate images that are more reminiscent of those in a lookbook. In the initial picture, the model merely stood straight. So, I decided to add a hint of the world view and specify a pose.

I used prompts "background + pose specification", "Inside a spaceship of the future", and "sitting", resulting in the image below.

 

 

In lookbooks, even if the outfit isn't entirely visible due to the pose, it's generally not a concern. One advantage of Midjourney is its ability to generate a background or chair that matches the outfit or the world view specified in the prompt.

 

Making it Suitable for E-commerce

For e-commerce, photos should be clear and informative. Typically, images should clearly showcase product details, texture, and colors, often against a uniform background like white.

The result of adding "only background" and "white background" can be seen here.

The top left is a coat, but it doesn’t provide a comprehensive view of the product. To capture the full outfit, I tried "Pan Down", which extends the image downward.

With the entire outfit now in frame, potential buyers can assess the full details of the product.

    To close

    • While I've generated images for various purposes, it was more challenging than I thought.
    • I suspect my lack of knowledge about angles and filters played a part. Though AI enabled the creation of amazing images, possessing more photography knowledge and techniques might lead to more targeted image generation. From now on, I'll be observing fashion images in magazines and websites with the concept of prompts in mind.
    • This time, I employed Midjourney's "Pan Down" feature to generate desired images. However, further editing using tools like Adobe Photoshop seems crucial for the future.
    • Apart from lookbooks and e-commerce, fashion photos serve other purposes like social media, magazines, advertisements, etc. Understanding the unique characteristics of each and translating them into prompts could make image generation even more engaging.
    • Building on this experience, in my next blog, I intend to create an actual lookbook or portfolio for "accelerando.Ai". Stay tuned!

       

      To conclude, I'd like to showcase some of the splendid images generated in this exercise.




      I'm eager to continue experimenting and enjoying this journey!

       

        Back to blog

        Leave a comment

        Please note, comments need to be approved before they are published.