This workflow allows you to choose a folder of LoRA and run a prompt applying one at a time in ascending, descending or random order.

It's great for when you have a folder of characters, for example, or want to see what your prompt would look like with each of your style LoRA applied.

I've included SDXL and FLUX versions but literally the only difference is whether it uses the CLIP loader so you can likely use it with whatever image model type you'd like.

Features

• You can apply static LoRA separately

• It will save in CivitAI-compatible format

• It will extract any trigger words from the randomized LoRA (using Comfyui-Auto-Trigger-Words)

• You can batch process using ComfyUI's built-in rools.

Source Material

This is a rewrite of the excellent Lora Swapper Workflow by @Catz. I cleaned up the formatting, switched to RegEx to correctly parse path names, and made some other layout and formatting changes.

How To Use

Get all the configuration set up, then set ComfyUI to queue up as many images as you'd like

Limitations

• You cannot customize the LoRA strength for the randomized folder when running the workflow in batches.

Required Nodes

(These are all available in the node manager)

• ComfyUI-Easy-Use (for miscellaneous stuff, you should really have this anyway)

• ComfyRoll (To send randomized LoRA through the pipeline)

• ComfyUI-DirGir (Randomization)

• ComfyUI-KJNodes (For various text extractions)

• ComfyUI-MXToolkit (dimensions grid)

• ComfyUI-ImageSaver

• Mikey Nodes (for prompting)

• SaveImageWithMetadataUniversal (for CivitAI-compatible files)

• ComfyUI-Lora-Auto-Trigger-Words (To extract trigger words from randomized LoRA)