Any ideas on if/when (image+text)->image models will be available?

Would be good to be able to iterate on images (keep this, change that etc).

The use of the seed looks useful but I'm guessing it has its own limitations.