Stability AI’s Stable Doodle Turns Even Bad Sketches into Fine Images

Stability AI has released a new AI image generator called Stable Doodle which converts sketches into images using their new Stable Diffusion model “SDXL 0.9” which provides “limitless imaging possibilities.”

Stability AI, an open source generative artificial intelligence company announced the launch of Stable Doodle on July 13.

According to their blog, the latest tool brings drawings to life and it can “dramatically enhance a range of industries, from education to creative design, to fashion and the arts.” The new tool was made available effective July 13, through Stability AI’s March acquisition – Clipdrop platform.

From sketch to artistic image

The new sketch-to-image tool which converts simple drawings into dynamic images does not only take sketches as input, but rather users can also add text prompt to improve the generation process and results. One can add “dog in cap.” Stability AI however highlighted a limit to the customization as the new tool can only support 14 styles of art.

“Stable Doodle is geared towards both professionals and novices, regardless of their familiarity with AI tools,” said Stability AI.

“With Stable Doodle, anyone with basic drawing skills and online access can generate high-quality original images in seconds.”

According to TechCrunch, an online portal focused on high technology and startup companies, Stability AI’s sketch-to-image tool is more precise when compared to the many tools available on the internet.

Looking at the sample image generations they shared on their blog and in a short video from their Twitter page, the new Stable Diffusion model is pretty impressive. And the quality of the generated image improves significantly provided the sketch had more detail.

Behind the images generated by Stable Doodle is a combined effort between the Stable Diffusion XL model and the T2I-Adapter, a condition control solution from Tencent ARC.

“By adding trainable parameters to existing large diffusion models, the T2I-Adapter enables inclusion of additional input conditions such as sketches, segmentation maps, or key poses,” according to the blog post.

The T2I-Adapter allows the pretrained SDXL model to understand the outlines of the sketch to be able to turn it into an image.

“The T2I-Adapter network is comprised of approximately 77M parameters. It offers additional guidance to pre-trained text-to-image (SDXL) models while keeping the original large text-to-image models unchanged,” they wrote on the blog.

Also read: Europe’s First Bitcoin ETF to Launch Amidst Renewed Institutional Interest

Democratizing image creation

Technology advisor Jeff Wenzel told Lifewire, an online portal which focuses mostly on technology information that the new image creator can become highly accessible to a wider audience if it has the potential to democratize AI image creation.

He said the tool would help solve the language barrier challenge that often limit users in achieving their desired outcomes.

“Sketching a doodle, on the other hand, taps into human intuition and creative expression, bridging the gap between individuals and AI algorithms,” he added.

He however noted that: “While Stable Doodle simplifies the image generation process, there may be a trade-off between accessibility and obtaining precise results.”

Accessing Stable Doodle

The Stable Doodle tool is available to use through their website, IOS (for Apple product users) and Google Play App Store (for Android-powered smart devices). While usage has been made free for the time being which grants users the ability to experiment with the image generation tool without any login (credentials), there are limitations to the number of images that can be created daily.

“The user-friendly interface is designed to be intuitive: users create a basic sketch with a mouse, choose an art style, and click ‘generate’. It’s that simple!” explained the company. But Stability AI is not stopping there.

“Moving forward, Stable Doodle will enable users to import a sketch. Further, we will include use cases for specific verticals, including real estate applications, for example.”

Industry competition

On July 13 Google updated their chatbot Bard to accept image input and save conversations much like its rival ChatGPT. However, for ChatGPT, image input is only available to Plus users and ChatGPT 4.

Another rival in generative AI, Anthropic released Claude 2, another AI chatbot similar to ChatGPT and Bard. Anthropic processes large block of texts. It looks like the race to produce the best AI tool is not stopping anytime soon with companies doing what they can to design products that comply with global regulations on security and privacy.

Stability AI is said to have raised about $125 million. The startup is valued at $1 billion and the team behind the company is working towards quadrupling their valuation within the next few months.

Image credits: Shutterstock, CC images, Midjourney, Unsplash.