Is SD making advance in prompt understanding and generation coherence, as the new Dalle 3 has made huge progress in it?

[lanyusan]

A general question for yasviel and sd's deep experts here.

Obviously Yasviel is a giant in the SD ecosystem as inventor of controlnet. Without it, SD would have lost at least half of its power. I am sure Yasviel knows more than us if SD's core developers, no matter they are Stability AI employees, academic researchers or open source contributors, are making progress in this front.

I am wondering if Yasveil or anyone could share some insight views. And I am also curious in what form would that be? An entire new layer like controlnet ? Some new models? Or pipelines etc. ?

Just a trigger question. Hope to ignite some interesting discussions.

Thanks.