how steerable and coherent is it with tag conditioning? can you do like a small foxgirl with red hair and blue eyes in the snow holding an ice cream? (via tags, that is)

can we have more precise control over the poses/actions? for example, Hatsune Miku touching the ground with her hand?

Is it possible to input a photograph, and have the model make the anime version of it?

Haven’t tried, but I have some ideas , probably worth to do some experiments

That is awesome. Can it only create artworks with women right now or can it do men too? Say like a pink haired boy with purple eyes? Or if not what about a brown haired angel girl with pink eyes?

Of course! the model is capable of generating guys

Can it create art works for a specific anime girl? e.g. Ayanami Rei is waving her hands and smiling at you

Yes it’s possible (but the model messes up the number of fingers sometime)