This has 4 Redux image inputs and there's a fast group bypasser to the left of those and you can enable/disable the ones you want or don't want to use. I'm using the ComfyUI_AdvancedReduxControl custom node with the StyleModeApplySimple node for each Redux which allows you to set the strength of them.

For Flux Force/SetClip device is being used to offload the clip models to normal RAM, and Detail Daemon Sampler is connected.

Redux seems to ignore text prompts (you can lower the strength to combat this), but I found the text prompt is important when coupled with an image input. For example I added 3 images from an Eladrin D&D character in her spring phase I'd been working on and for the 4th image I added a pic of a woman wearing a red beret.

I set the top image to High, 2 and 3 to medium and the last image of the woman with the beret as low and added the prompt "an elven warrior with long green hair and wearing a red beret stood in a forest" and it created an image of the elf wearing a red beret.

I then removed the 4th image and replaced it with one similar to the first 3 and no beret on the output.

I then swapped the images to ones of her in her winter aspect, had the woman wearing the beret on the 4th and it worked as expected. Removed the prompt but left the image and no beret again. So seems like the prompt is working as long as there's a visual input to go with it.