Fly With AI Logo

Fly With AI

Stable Diffusion Seeds Explained + Settings

Stable Diffusion Seeds
On This Page

Today I am going to cover how stable diffusion seeds impact your images and the best way to use them.

This will help you in your experimentations with other parameters and help you find images similar to those that you like.

Let’s dive right in:

On This Page:

What Are Stable Diffusion Seeds?

A seed in stable diffusion is a number assigned to the starting noise. This starting noise is used to create the image along with your other parameters such as the prompt, sampler, and CFG scale.

Stable Diffusion Seed Setting

Changing seeds has a big impact on what your images look like because the starting noise heavily dictates the final outcome of your generated images.

How to Use The Seed Setting

Seeds can generally be set with two options:

  1. Random: Random will automatically choose a random value for the seed each time you generate images.
  2. Recycle: What I mean by recycle is using the same value or number for the seed, so the starting noise is the same.
Seed variation setting

I have a few recommendations for when to use a random seed or recycle the same one:

  1. Random: If you generally like the outcome of an image because of the prompt and other settings you are using, then generate images using different seeds until you find the best outcome.
  2. Recycle: If you are experimenting with changing different prompts or parameters, you should keep the seed the same. That way, the changes you see in different image generations are due to the other parameters changed, not because the seed had changed.

This is everything you need to use seeds effectively.

Next, I am going to cover a couple of common questions with using this setting:

What is The Seed Range?

Pytorch, which is used to run stable diffusion, has a minimum and maximum range for the values that a seed can be:

Minimum: -9223372036854775808

Maximum: 18446744073709551615

Anything beyond this range will return an error.

What Does Stable Diffusion Seed -1 Mean?

Stable diffusion seed -1 is something that you will see in your settings, indicating that a random seed is being used each time you generate an image with a value of -1 of the previous value.

additional seed settings

For example, the first image would have an initial value of 102, the second image would have a value of 101, and the third image would be 100.

Comparison of Different Seed Values

I am going to show you generated images using different stable diffusion seeds to show to the influence that it has on the final output.

Image With 3520749160 Value
3520749160
Image With 3767920236 Value
3767920236
Image With 396708388 Value
396708388
Image With 940910824 Value
940910824
Image With 45546546 Value
45546546
Image With 1858502499 Value
1858502499

Prompt: castle, ornate, beautiful, atmosphere, vibe, mist, smoke, fire, chimney, rain, wet, pristine, puddles, melting, dripping, snow, creek, lush, ice, bridge, green, stained glass, forest, roses, flowers

Each image still follows the prompt syntax correctly but has some distinct differences in how it is interpreted.

For example the castle with the 396708388 value has white paint, also some of the images have more colorful roses than the others.

I’d Like to Hear From You

Now that I have covered everything regarding using stable diffusion seeds.

Let me know if you have found a good way for choosing between random or recycling seeds, or any other tips, by leaving a comment below.

Leave a Comment