Maybe the private/paid ones can do more artstyles but the free ones kinda' have their own intention to force people's inputs into looking like Pixar, movie posters, video game concept art, anime fan art, etc. It's not actually that diverse, or at least not diverse without the user inputting like dozens of specific instructions to try to guide it away from being generic as fuck where the machine's most comfortable.
Maybe the private/paid ones can do more artstyles but the free ones
I'm only familiar with locally hosted open source models, which have way more ways to control the output from LORAs to controlnet and things like IPAdapter.
Maybe the private/paid ones can do more artstyles but the free ones kinda' have their own intention to force people's inputs into looking like Pixar, movie posters, video game concept art, anime fan art, etc. It's not actually that diverse, or at least not diverse without the user inputting like dozens of specific instructions to try to guide it away from being generic as fuck where the machine's most comfortable.
I'm only familiar with locally hosted open source models, which have way more ways to control the output from LORAs to controlnet and things like IPAdapter.
They aren't running it on their own PC, this is being put into some web interface.