“An astronaut riding a horse as pencil drawing”. This and many more images can be easily crated on the fly in various versions from nothing more than a piece of text:
Don’t like the style? No problem, just ask for “An astronaut riding a horse in a photorealistic style”. That’s all it takes with new OpenAI DALL-E 2 model which can literally turn words into images.
But not only that. You can mask an existing image and ask for modification and then model will automatically adhere to existing image style and fulfill your with. Let’s take for example a following image:
… mask the poor creature …
… and ask model to replace if with you different specie.
And say you want something like koala riding a bike? Just say the word and you get a bunch of different koalas to choose from.
There’s no sacred image for DALL-E 2. It can automatically match the style of any image it edits. Don’t like the Mona Lisa’s hairs? Say no word.