There is a certain frustration that comes after generating AI artwork ; it feels like a lottery, some prompts will return great images (but with a little defect here and there) and then there are many failed attempts, where the images are just not good, partly out of frame, or simply with deformities, or artifacts (purple blotches, weird eyes, weird hands, etc ...)
Surely, this is normal, and AI artwork is still in its infancy, or at least early ages ; it is still impressive. but as an artist, you need a little more control than just changing a word here and there, and pressing 'run' or 'launch' and hope for the best.
In the coming year, there probably will be some kind of a sketch application, and AI will then improve the images (man made) ; instead of asking the AI to do everything, this is probably what will be the definitive step forward.
controlNet is going in that direction, by helping an AI or strongly suggesting a pose to start with.
It can be based on a sketch. And then there is invokeai, which offer an environment for the creative minds.
No comments:
Post a Comment