Can it be trained for style ?

[scarbain]

Let's say we give an image style (for example a painting of Van Gogh) as control image and the target must match that style but following the prompt.
Could it work ? Or does the informations the control images carry are too much based on the position of the pixels inside ? (because all the given examples from the repo are based on positional information)

[geroldmeisinger]

Composer: Creative and Controllable Image Synthesis with Composable Conditions