Comparing 2 Models in Automatic 1111

Learn how to compare two models in Automatic1111 so you can evaluate output quality, style differences, and workflow fit more effectively.
Hero image for Comparing 2 Models in Automatic 1111: Learn how to compare two models in Automatic1111 so you can evaluate output quality, style differences,...
Hero image for Comparing 2 Models in Automatic 1111: Learn how to compare two models in Automatic1111 so you can evaluate output quality, style differences,...

There are times when you will need to compare models. You may need to determine which one is the best for a prompt you are working on for a project. It may also be beneficial when you begin training and merging models. With that in mind let me show you real quick way in Automatic 1111 to compare 2 models.

Step 1: Start up Automatic1111

Select Auto1111 and hit select.

Screenshot showing training and merging models

Then on the next page make sure you are on a Medium or Large box. I tested it on both. Medium works but large is faster.

Screenshot showing 2023/12/Step1a

Step 2: Change the Settings

Let’s let Automatic1111 have more than 1 checkpoint (model) available at a time this will save on load and image generation time. Click on the Settings tab first. Then on the left hand side click on stable diffusion. Then you’ll want to change where it says 1 to 2 for “Maximum number of checkpoints loaded as the same time.

Screenshot showing ettings tab first

Don't forget to click on Apply Settings.

Screenshot showing Don't forget to click on Apply Settings

Step 4: Set Up our XYZ Plots

Click on the txt2img tab next.

Screenshot showing c/79/65/7965787d-54cb-4211-8169-3d231a1bb6f8/content/images/2023/12/image-3

Scroll down to Script. Click the dropdown error and select X/YZ Plot

Screenshot showing g-image-card"> Scroll down to Script

With XYZ plot selected it will give you access to a few more fields. In the X type field select checkpoint name. Our  X Values we will select two of the models we want to compare.  Then uncheck draw legend because it can sometimes throw errors. Since we have two models we know that the model we select first will generate first and be on the left hand side and the second model will be on the right hand side.

Screenshot showing throw errors

Now compare the two models. I am trying to generate a cute mouse eating a blueberry while wearing a Santa hat with snow next to a window. You can see which model followed my prompt so I know that of these two models for this type of image I shouldn't use the base model.

Screenshot showing wearing a Santa hat with snow next to a window
Screenshot showing s type of image I shouldn't use the base model
About the author
Adam Stewart

Great! You’ve successfully signed up.

Welcome back! You've successfully signed in.

You've successfully subscribed to RunDiffusion.

Success! Check your email for magic link to sign-in.

Success! Your billing info has been updated.

Your billing was not updated.