Table of Contents
What is the Seed parameter in Midjourney?
In Midjourney, Seeds are numbers that are used to create a visual noise to initialize the generations of an image or a set of 4-quadrant images.
The seed determines the starting point for the generation of an image, which is created from a field of visual noise similar to television static.
By default, Seed will not be added to any prompts. If the Seed is not specified in a prompt, Midjourney randomly generates an image/ a set of 4-quadrant images.
For example, I run the same prompt twice without the Seed parameter:
By adding Seed parameter with your old same prompt, you can bring similar as well as identical images.
I added the Seed value of the above first image along with the same prompt and ran it a couple of times. The results are:
The results are not just similar but identical. The Seed parameter works well with the Midjourney version v4.
How to Find a Seed number of my Midjourney Images?
To use the Seed parameter, you need to know the Seed number of your pre-created Midjourney image.
To find the Seed number, hover over the image you have generated. Then, click the “three horizontal dots” located in the upper right corner of the image.
Next, click the “Add Reaction” to find and click “envelope” icon.
By doing so, Midjourney Discord will DM your Seed value.
Now, you can use copy the Seed value and use it in your prompt with the parameter “–seed”. Always remember that you can only copy the Seed value of your own jobs.
If you want to know the Seed value of jobs created by fellow Midjourney users, you can ask them nicely.
Seed Parameter and Midjourney Model Versions
The Seed parameter works with all Midjourney versions, such as v1, v2, v3, v4, test, and testp.
However, you can get identical images only with version 4 and niji. The usage of Seed parameter with other Midjourney versions will result in only similar images, not identical.
See the image below for an example:
These images are created with Midjourney version of 3.
Midjourney “–seed” Parameter with Other Parameters
The “–seed” parameter can be used with other Midjourney parameters according to the need of users. The input image for this experiment is:
–stylize
Yes, there is a slight difference lies between the seed image and this image. When I run this same prompt with a different style value in Midjourney v 3, the result is total chaos and different.
Note: Stylize value in Midjourney v3 is “625 to 60000”, whereas, in v4, it is “0 to 1000”
Aspect Ratio
I was not expecting the result of using –seed with the –ar parameter.
As you can see, there are no similarities between the seed image and the above two images!!! The color composition and the place and appearance of the subject have a difference.
Subject Swap
Let’s do the subject swap now.
Since the subject modification is a higher form of change in prompt, there is much difference here. No surprise.
Like this, there are many possibilities and impossibilities when combining the –seed parameter with other parameters. To find out, kindly do your own experiment based on your requirement.
What is –sameseed?
The –sameseed parameter in Midjourney is same as –seed. The difference lies in the noise. The –seed parameter triggers a single large random noise across all four images in the initial grid, whereas the –sameseed triggers the same starting noise across all four images.
Hence, the –sameseed can produce more similarity over –seed.
The same –sameseed parameter is only compatible with Midjourney model versions 1, 2, 3, test, and testp.
Since these models did not create identical images when –seed is applied, the –sameseed could be the solution for these versions.
Conclusion
Each image generation process of Midjourney will have a randomly generated unique Seed value.
By using the Seed value along with the same prompt, you can generate similar or identical images again.
The usage of “–seed” parameter preserves the consistency of the results even if you change the subject or add any elements on the whole. You can use –sameseed parameter in Midjourney versions 1,2,3, test, and testp to bring more similarity.
FAQ
1. Difference between Midjourney seed parameter and remix mode?
Both the Midjourney seed parameter and remix mode are used to maintain the same consistency of the image while modifying the prompt.
The difference is that the Remix mode can only be applied to your current or future images, whereas the Seed parameter can be applied to your pre-created, current, or future images.