Stability AI Ltd. is adding to its arsenal of generative artificial intelligence tools with the launch of Stable Doodle today.

The company said it’s a sketch-to-image tool that works by converting simple sketches into more dynamic images, and it’s aimed at everyone from creative professionals to hobbyist artists.

The launch of Stable Doodle builds on Stability AI’s flagship open-source model, Stable Diffusion, which is a generative AI model that creates images based on text prompts only. The company isn’t only about images, though, offering an open-source model called StableLM that generates text and code.

Stable Doodle is a new tool within Clipdrop by Stability AI, and is designed for use in industries such as creative design, fashion, education and the arts. It’s being made available for free, alongside the latest version of Stable Diffusion, SDXL 0.9.

The company said Stable Doodle can be used by professionals and novices alike, and requires no familiarity with AI tools. Users can simply outline a basic sketch and upload it, and the model will transform it into a professional, cartoon-style animation, an origami model or even a photorealistic image, depending on the user’s choice.

Stability AI said in a blog post that the tool is meant to save developers, designers, illustrators and other professionals time, enabling them to maximize their efficiency. For example, a designer could sketch out a basic idea for a client’s design and then use the tool to create various compelling options. It can also be used to create materials for presentation decks or websites, logos and more.

Under the hood, Stable Doodle combines the same image generation model of Stable Diffusion with a powerful T21-Adapter, which is an open-source condition control algorithm created by Tencent Holdings Ltd.’s applied research center and allows for more precise control over AI image creation.

By adding trainable parameters to Stable Diffusion, the T21-Adapter allows to input additional conditions on the design of their images, including sketches, segmentation maps and key poses, which are frames used to denote the important movements taking place in a sequence. It supports multiple models of input guidance, giving users an enhanced level of control over the entire image generation process.

With Stable Doodle, the T21-Adapter offers supplementary guidance that enables the Stable Diffusion model to understand the outline of sketches it is shown. Images are then generated based on both prompts and the outlines defined by the sketch, Stability AI said.

Holger Mueller of Constellation Research Inc. said Stable Doodle is another sign of the astonishing progress of generative AI, which was virtually unknown this time last year but is rapidly becoming an essential tool for thousands of professionals.

“Nonetheless, it’s a reminder that some companies may want to pay attention to what’s under the hood of these generative AI offerings,” the analyst said. “In the case of Stable Doodle, its most innovative component is the open-source T21-Adapter, which was created by Chinese tech giant Tencent. Still, the results are very impressive.”

To get started with Stable Doodle, people can visit Clipdrop by Stability AI on its website or download the mobile app for iOS and Android. They can get started without any login, create a basic sketch with a mouse or their finger on touchscreen devices, select an art style and generate an image.

Images: Stability AI