What this workflow does

👉 This workflow is an adaptation of a couple of my other nodes.  This one just takes 4 images that get fed into the IPAdapter in order to create an image in the style and with the color of the images.  

In the example images I loaded 4 old time Santa and Christmas images in the 4  Style Image boxes.  Then I described to the postive prompt what I wanted to see in the output image node, trying to use what I provided in the Style Images. If I get something in the output that I don't like, I add it to the negative prompt. But you could load any 4 style images you want, they don't have to be christmas related.

This workflow should be able to generate just about any output the model can support. If you really need to, you could add a lora input  just after the load checkpoint.

How to use this workflow

👉 Find 4 images in the style and color you want to replicated.  Describe in the positive prompt what you see in the Style Images and anything you want added.  Put in what you don't want to see in the negative prompt.   Click Queue Prompt in the control panel.

Tips about this workflow

👉 Try to aim for similar styles and themes you want to see in the final image. This is like how you use the ipadapter in automatic1111, but you can put 4 Style images.  You can change the checkpoint to one you have on you system. The weights for the images can be changed in the Encode IPAdapter lma node.  You can see the progress of the ksampler just over the save image node.

This is set up to use sdxl models right now. You have to change the models over to sd1.5 to use those models in the checkpoint. This includes the load clip vision node and the load ipadapter model nodes.