This is a straightforward workflow for the new SD3.5L model using 3 loras that you can adapt and expand to fit your needs. Currently setup for photorealism with a cost of 1 token per image.