if you have time, i suggest if you have a nice image (with a turbo lora) deactivate your hyper or lightning lora and only change parameter below!
for flux and qwen you can use:
euler / simple ~40 steps and CFG ~2-3
-> you will be surprised i hope ;)
(esp. qwen-edit and flux-kontext)
AND dont use GGUF only if your VRAM lower than 12GB
-> USE FP8 version, even if some GB more than your VRAM, it is 2 times faster than GGUF and quality nearly Q6-Q8