This lora was training on telephoto image pairs and helps zoom in really close to the subject. At strength 2.0 the effect can be pretty strong sometimes as seen in some examples. For a milder effect you can just use +1 str. H3 is a very capable model and while correct prompting can get you pretty far, sometimes a niche lora can help out.