>>106953507
Oh, you know, like the only model that properly understands both what photos look like (which even the latest API models still struggle to) and in terms of open source, where the objects go in relation to each other, proper spatial awareness, as well as coherence regardless of how simple or complex your prompt is.