New AI design model SD-V3

overview

Starting in June 2024, we have added Stable Diffusion3 to the models that can be used in AI design.

Specific benefits will be introduced in the "Use Cases" section later in this article.

Please note that the image limit for the previous Stable DiffusionXL and the newly added Stable Diffusion3 are each counted separately.

Please refer to the pricing page for the maximum limits for each model.

How to use

① Model switching

• From [Account Settings], select [Model to use] > [Stable Diffusion3] and click [Change].
  *It is also possible to revert to the previous [Stable Diffusion XL].

②AI-designed UI

In Stable Diffusion 3, the UI for AI design has been changed.

• The previous Stable Diffusion XL [Image Style] is no longer available. From now on, please specify the expression method such as [Line Art] or [Pixel Art] directly in the prompt.
• Instead of [Image Size] it is now [Aspect Ratio].
• You can now select the output format.

Use Cases

①Improved prompt reflection accuracy

The text you type is now more accurately reflected in the image, especially when it comes to detailed instructions about the number of people in the image and their clothing.

prompt:

Design, entertainment, democratization of reality, various perspectives and uses, high fashion editorial photography, background color is white . The three women are wearing blue, white and red outfits respectively ( left woman: blue, middle woman: white, right woman: red )

②Improvement of character generation accuracy

We've improved the accuracy of generating text within an image, making it easier and more effective, especially for logo design.

Please note that it is not possible to generate characters other than the alphabet, such as hiragana or kanji, within an image.

prompt:

[Logo name] "Stylish Shirts Co."
[Design style] Modern and simple [Font] Sans serif, bold [Color] Deep blue and white [Icon] Small silhouette of a shirt [Layout] Icon on top, logo on bottom, surrounded by a circle [Background] White

Stable Diffusion XL

3) Improved accuracy of additional image correction

It is now possible to adjust the strength of the generated image when it is corrected again by AI. This makes it possible to make effective adjustments while maintaining the composition and pose to a certain extent.

*This is a function to edit the generated image. Currently, there is no function for users to freely upload images.

Original prompt:

Line art, cat, full body

Additional correction prompts:

Cat, realistic