in you're opinion what text to image AI should explore next?

would prefer something I can run myself / self host.

I've used midjourney and dabbled in stable diffusion.