[Stable Diffusion] Verify the difference between 4 Resize modes that can be selected with img2img!

• Which Resize mode should I choose?
• I always make the right choice...
• What difference does the generated illustration make?

We will answer these questions for you.

Verify the difference between 4 Resize modes that can be selected with img2img!

[Stable Diffusion] Verify the difference between 4 Resize modes that can be selected with img2img!

When generating an illustration with Stable Diffusion's img2img, you need to choose one of four Resize modes. If you make a mistake in this choice, "Oh, the result will not be what you expected." . "

In this article I will explain :

• [Stable Diffusion] Detailed explanation about Resize mode of img2img!
• Verification of the difference between the four Resize modes
• Setting method if you want to use another Upscaler
• If you want to generate illustrations more smoothly...
• How to easily manage a large number of generated illustrations

so if you read to the end, you will understand the difference between each Resize mode, and you will be able to get the illustration you want.

If you want to know how to use img2img in general, please read the following articles as well. It is explained with an illustration that was actually generated.

>> [Stable Diffusion] Explain how to use img2img to generate images from images!

[Stable Diffusion] Explaining the Resize mode that can be selected with img2img!

• Just resize
• Crop and resize
• Resize and fill
• Just resize (latent upscale)

Let's actually verify how the illustration is resized using these four Resize modes.

The conditions for verification are as follows :

• Original image size: 512 horizontal x 512 vertical
• Size of the illustration you want to generate: Horizontal 1024 x Vertical 512
• Denoising strength: 0

We will do the work of doubling the size horizontally while maintaining the pattern of the original image.

Just resize

As the name suggests, it is a way to change the size as it is. In the previous example, it's the type of "stretching with force".

Let's see the actual generated result :

[Stable Diffusion] Explaining the Resize mode that can be selected with img2img!

The top and bottom are protruding because the aspect ratio is kept as it is. It is a mode to be careful of because the composition changes greatly.

This is a good choice if you want to emphasize something in the center.

Resize and fill

In this method, the size is changed while maintaining the aspect ratio, and the blank area is newly drawn.

In this way, the newly generated blank areas (orange squares on the left and right) are drawn by Stable Diffusion in its original form (Denoising strength: 0 this time, so you can not draw freely, just stretched only).

If you increase the denoising strength, Stable Diffusion will draw with flexibility, which is quite convenient.

Just resize (latent upscale)

This is the same idea as the first "Just resize", and resizes while changing the aspect ratio. However, only the Upscaler when generating illustrations is a different version.

Different Upscalers will generate different illustrations, so choose the one you like.

In this verification environment.

Let's change the item of "Upscaler for img2img" from "Setting" → "Upscaling".

After selecting, don't forget to press "Apply settings" and "Reload UI" at the top of the screen.

There is no particular change on the screen, but the one selected earlier has been changed to the Upscaler used in "Just resize".