I've not had much luck creating this concept in SD 1.5 so I generated a bunch of images in Flux and trained a 1.5 LoRA on it.

Try at least a LoRA strength of 0.7. Sometimes you might get floating tissue boxes - I haven't found a way around it yet though I may update this model if I find a way around it. Maybe you can try one of the sample images where the girl is holding the tissue box in the center and use it as a controlnet image.