>>2746750Can you elaborate? I'm not familiar with artificial-v3, you mean Anything-V3?
In general, the different models have different prompts that work for them. For instance, NAI is weak at drawing eyes, so you have to give it focus, but Anything requires no special attention.
Tags work differently across models as well. "masterpiece" is a nice trick on NAI but has little effect on Anything.
Here are common negative prompts that I use for Anything-V3:
watermark, artist signature, male, simple background, extra fingers, missing fingers
The other negatives, "error", "extra digit", "fewer digits", "bad anatomy" and such have little to no effect when used with Anything.
My favorite samplers, for pretty much any model, are Euler a and DPM++ SDE Karras. I tend to stick with 7-8 CFG scale and 28-40 sampling steps.
