You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Sounds odd. You'll often get something that isn't quite right, but as long as it's "close", try running that image again (using img2img): increase the resolution slightly, and set a prompt-strength of around 0.3, generate, and often the AI will improve results. Also, be sure to run the first generation at the resolution that matches the model. For SDXL, there are a list of supported resolutions, but anything near 1024x1024 may work fine.
That really depends on the model. I found negatives work on simple things and even then its better to say - no smoke. But destroyed hands, extra limbs etc, has more to do with the lack of training than the engine. Large images will often give extra limbs. But some models are just trash when it comes to hands and limbs, forks, legs, etc
I put deformities as negative but everything is deformed, hands, faces, legs, limbs, every part.
How to do it right?
The text was updated successfully, but these errors were encountered: