After several tests the best setup for me is:

• Base-Model: Stable-Diffusion-3.5-Lage

• Lora-Strength: I prefer 0.8 for images (but depending on the image you can try different weights)

• VAE: Automatic or None (depenting on your base model it might be a good idea to change it too)

• Sampler: DMP++2M is getting pretty good results, but depending on the king of image you can try different sampler

• Sampling-Steps & CFG: The examples have been created with 20 steps and a CFG of 3.5

You can scroll down and see some example images you might want to use by clicking remix so you can just transfer the creation setup of that spec. image to your creator.