>>2826068I used that zoom slider lora to get more space in the canvas but nothing else. The base gen was also in 768x768 instead of 512x512.
I also find that writing a small description of the image before the actual tags usually helps steering the overall composition in the direction of what I had in mind. Just try to avoid a lot of redundancies or contradictions and it should work fine.
Foxbox if you're interested:
https://litter.catbox.moe/a8wt6x.png 