Comparing "img2img" vs. "ControlNet Depth Mode"

You may be wondering what exactly the differences are between using a Reference Image with and without ControlNet, so let's take a look at that!

Here we are using a simple prompt to make a potion on our SD 1.5 Potions model

Next we will select one of the outputs as our reference image and generate new images using only img2img


These are interesting and cool results. but what if I wanted to maintain specific features of the image composition?

Let's activate the depth mode and compare our results

As you can see below, img2img creates variants with differing shapes and adornments, while depth mode maintains the shape and proportions

This is just one of the many ways we can explore how composition control provides ultra consistent asset creation.