I've had some fun with pixray today, which can generate stylized (especially the pixel-art thingie here is cool, but there are different flavors) images from a textual description of something. That's nothing new in general, but the quality of output is really something, I think. It's fascinating to see what the algorithm ends up with, and the interface allows you to see several intermediate steps so there's a bit of evolution to be seen here. Here are some example outputs and the text that generated them (using the same text will likely yield different results though, as the image generation is not idempotent):
a whale falling from the sky
two hamsters fighting to death
the code will eat all
run before the troll catches us
santa slaughtered everyone
a brain in a box covered in stardust
rainbows and sunshine and pancakes
hunt showdown