The Google Research team has unveiled StyleDrop, a neural network that generates images from text descriptions. There are already many of them, you might say. Yes, but the main feature of the new service is that it generates images that exactly match the given style.
StyleDrop generates any images by text description. To set the style of the result, it is enough for the algorithm to show a reference picture — whether it is a work of a famous artist or a children's drawing. All text prompt generations will follow the style of this image.
The creators of the service suggest using it, for example, in creating an entire brand concept based only on the logo. In addition, StyleDrop does a great job of generating text on pictures.
While the product is still awaiting real-world testing, we are eagerly looking forward to its release.