I built a workflow in comfy UI, it loads images and tags them, it combines the tags into a prompt then gens the result without doing a i2i, its just using tags from the images and making a t2i gen.