It really comes down to the prompt-cfg (Classifier Free Guidance) setting relationship. cfg = 3.0 (I usually try a range of cfg settings, depending on the model, from ~2.4 to 4.0) #FLUX1FP16 using ddim_uniform scheduler / dpmpp_2m as the sampler. On this run, >3.3 was not good; lots of artifacts.

Excellent, but sorry don’t know what you are talking about🤪. I use Ideogram app.