Trained for an edit workflow using Qwen Image 2.1 Turbo model and its 8 steps, you supply the hole, and this LoRA should fill it...or at least that was the intention. My first attempts were very successful, but then the more I fussed with the prompt, the worse it got, so keep it simple. Leave enough room for the top to get in there, or specify that the camera is changing, or outpaint your start image.

This is a non-commerical model, so this one is free. Enjoy and share below with Add Post

I included snapshots at 800 steps and 900 steps if you have problems with 850 steps main file on here, and let me know what you find.