Increase SD 3.5 steps above 30

Increasing step counts often increases detail, prompt adherence and can decrease the likelihood of warped images. Venice still doesn’t offer any models that aren’t specifically tailored to realism and while SD is more versatile generally a large step count helps detail and enforce anime or 3d animated art styles. Flux can do anime images but gives everyone childlike faces especially women despite explicit prompting. None of the other models are remotely capable of non-realistic art style generations, especially something like anime.

Same Prompt examples on SD 3.5 and Flux Standard (the Highest quality SFW one)

Flux

SD 3.5

SD 3.5 with the same prompt before the change with a step count of 43

Neither SD image is good but like I said the model struggles with unrealistic art styles.

Either way SD was the only model Venice offered that didn’t focus so heavily on realism (flux so heavily focusing on childlike anime appearances) and that was due to legacy implementation from older SD models, and was more creative. It definitely isn’t effective when prompting for non-realistic art styles with previous SD models being better. Due to that it’s even more a matter of luck if you’ll get a good generation with step counts being higher to a degree making it more likely, thus usable.

With all that in mind please return the step count to 50 or add access to the models I and another user suggested a couple of weeks back to give Venice’s image generative models greater capacity/variety.