Stability AI, the start-up that has already created Stable Spreadlaunched a new tool for creating images by artificial intelligence.
This time, you will be involved a little more than by a simple description of the requested image. The prompt here is a sketch made directly by a Stable Doodle tool. It will then take him a few seconds to offer you three interpretations of scribbling in more or less artistic or realistic images, quite stunning.
How Stable Doodle works
In a blog post, Stability AI introduces Stable Doodle as “ geared towards both professionals and novices, regardless of their familiarity with AI tools “. After a quick test, it is undeniable that the grip is fast. All you have to do is go to Stable Doodle on the clipdrop.co site. The page that opens then presents a white square and even more basic drawing tools than those of Microsoft Paint to draw a quick black sketch.
Your work of art is finished, you can embellish it with a quick description in prompt, and above all choose the style of art requested (photographic, origami, isometric, comic book, etc.), then launch the generation. Stable Doodle then offers you three images that meet the criteria. It is not yet possible to import a drawing into Stable Doodle, but Stability AI has indicated that this possibility will be added soon. In addition, the list of styles offered, currently limited to 14, should also be extended.
While Stable Doodle has a free version, you have to pay a subscription to generate more than a few images per hour.
A cool feature, not really a revolution
While Stable Doodle is fun and works pretty well, Stability AI isn’t reinventing the wheel here. The generated images certainly follow the initial sketch quite faithfully, but they cannot do without a written prompt, and the result becomes more convincing with a complete and descriptive prompt. No huge change compared to other image generation tools, then. But the question of copyright in artificial intelligence could on the other hand become even more complex.
By the way, as Stability AI recognizes, similar tools already exist, but what sets Stable Doodle apart is its accuracy and fidelity to the source material. The target here seems to be designers or graphic artists for whom a purely textual description is sometimes not enough to transcribe the vision. The objective, of course, is to monetize Stable Doodle quickly, because without being on the street, Stability AI has not benefited from the same type of funding as OpenAI for example.
Sources: TechCrunch, Stability AI
2