These very powerful machine learning models allow users to generate incredibly photorealistic digital images, from more or less precise descriptions in natural language, known as prompts.

But there are also image-creating tools that are less oriented towards photorealism, and focusing on a fun, playful approach instead.

This is the case of Scribble Diffusion.

This web application, which can be accessed free of charge, allows users to convert simple sketches or drawings into more or less realistic paintings.

The idea is to make a drawing with a finger, a stylus or a mouse, with a brief text description and then obtain a much more artistic equivalent in one click.

Of course, it’s possible to “touch up” the image afterwards too for a result more in line with your expectations.

Scribble Diffusion is an open-source project and its code is available on GitHub.

The programme is based on a brand-new neural network structure called ControlNet. Here, artificial intelligence interprets the drawing to then generate the closest possible “artistic” image.