New Feature! Testing Image Prompts Offered in the Store! Logo Design Edition

February 29, 2024

New Feature "Store" Release!

I am Ushiyama from OpenFashion.
This time, we would like to look into what can be done using the "Store feature" added to MaisonAI on 2/14.

60 types of prompts (instructions to AI) are all free! The new feature of the AI utilization support tool "Maison AI" accelerates AI utilization in the fashion industry - now also supports multiple languages including English and Korean.
*written in Japnanese

General tasks, fashion-specific tasks, and image generation are now provided across three categories, each with about 20 prompts.
Although it's called "Store," all the prompts are offered for free.

First, let's see what "Image Generation" was like before the Store was introduced.

Filling out a blank field with English words on your own can be confusing.
Unlike AI agents or templates, the hurdle is that you have to write in English, making it difficult to generate images.

Testing Store's Image Prompts: Logo Design

Now, let's actually try out the image prompts available in the Store!

Here is a prompt that helps create logos.
Instructions for centering the image on the screen and specifying the atmosphere of the logo are already set.

Changing the Motif

Let's try changing the words.

An eagle has properly appeared! The quality looks good enough to be used as an icon image directly. Since there's only one change, it's very convenient.

Next, I changed it to "torso with jacket." It turned out to be quite stylish.

Changed it to "man wearing brown glasses."
The word "brown" seems to have been canceled out by the latter part of the color instruction, but it looks ready to be used as a profile image.

Can text be included?

Let's change the word to "pineapple."
Since it's a logo, can it actually draw a design with text?

Text has appeared...! However, unfortunately, there are spelling mistakes. Most of the logos have the wrong spellings.
At our MaisonAI, we are using Stable DiffusionXL. I heard that the accuracy of text input has been improved, but it seems there are still some dubious areas.

It was impossible to produce text in Japanese.

As an idea for a text-included design, it seems usable, but it seems difficult to use as it is.

However, as an "Icon Generator," if you replace words and color instructions, you can quickly use the prompt, so please utilize it!

Summary

It was good that we can now quickly try image generation from where there were no prompts at all.
However, there seems to be room for improvement.

Challenge: Users don't know where to change

If you can read English, you can pin down where to make changes, but for those who cannot, there are still parts that are difficult to use.
In the future, by refining the syntax of prompts, we hope it becomes intuitive to understand where to make changes, so we look forward to improvements!

Challenge: Cannot input in Japanese

Writing in English can be a hurdle for users. For slightly difficult words, you may have to input with a translation tool in hand.
This part too, we will continue to develop so that it can be used in Japanese in the future, so stay tuned!