Stability AI sets the painting threshold to 0!

1a4bbec3a7380873e0a7b6f579d4da4d.gif

Author | GPT-4

Responsible editor | Tangmen leader

Produced | "Intelligent Realm", a column created by AIGC

b64e383a53a94154279630bfa962b948.png

editor's note

"Smart Realm" column update: Stability AI is new!

Its acquired Clipdrop has released a new Stable Doodle tool. My most intuitive feeling after using it is that it has lowered the threshold for painting.

Before Stable Doodle, the most painful part of using various AIGC drawing tools was to write prompts that were very detailed and did not look like human words. I think this is also a pain point for many people, otherwise there would be no such thing as ArtHub.ai. The Prompt ideas website is born.

87fc018cbfe1d53ec0249ec92073a9f3.png

Prompt ideas

The biggest change brought by Stable Doodle is the update of the Prompt pose:

You draw a sketch casually, and then match it with the extremely simple Prompt, and a relatively accurate and extremely delicate drawing is born.

It is very simple to use Stable Doodle, just open the https://clipdrop.co/stable-doodle webpage, no need to log in to the account (you will be prompted to log in after multiple uses), and draw directly on the sketch, supplemented by a simple prompt. .

For example, as a senior excrement shoveler, I want to draw cats:

e1b0b5fc67819e3850748afccb01b36c.png

Seeing the diagrams generated by Stable Doodle, my biggest feeling is: my sketches are not worthy of...

Except for the title and "Editor's Note", all content in this article was created using AIGC. After sending the same prompt to ChatGPT (GPT-3.5), GPT-4 and Microsoft New Bing with different content, I finally adopted the content of GPT-4, mainly because of the logic and accuracy of the content it generates All are better in comparison. At the beginning, the ones generated by New Bing looked very professional, but after careful screening, I found errors, as follows:

d8eed6ee567045dfd35151df981303b4.png

Next, let’s take a look at how GPT-4 reported Stable Doodle. If you find an error, please leave a message and let me know.

18fa29b2ff2e29c9303adcb1b8d65251.png

GPT-4: The Birth and Application of Stable Doodle

Stability AI company recently launched a new tool, Stable Doodle, which is a tool that can transform simple drawings into dynamic images, providing professionals and enthusiasts from all walks of life with unlimited image possibilities.

Stable Doodle Release and Features

Developed by Clipdrop and Stability AI, the tool could dramatically improve efficiency across industries ranging from education to creative design to fashion and the arts. Now anyone with basic painting skills and online access can produce high-quality original images in seconds.

Stable Doodle's user-friendliness enables designers, illustrators, and other professionals to save valuable time and maximize efficiency. They can immediately implement the mind maps drawn into their work, create designs for clients, contribute material to presentations and websites, and even create logos. The possibilities are endless and fun to use.

For example, here are some examples using Stable Doodle:

9ac3f333d673649bcd805d5bb9ad2285.png

Tip: Comfortable chairs, 'isometric' style

c8db8a92d47212e6ffb5e3be722819e3.png

Tip: cat in denim jacket, 'digital art' style

How the Stable Doodle works

The Stable Doodle combines Stability AI's Stable Diffusion XL advanced image generation technology with the powerful T2I-Adapter. T2I-Adapter is a conditional control scheme developed by Tencent ARC (license), which allows precise control over AI image generation.

By adding trainable parameters to existing large-scale diffusion models, T2I-Adapter is able to incorporate additional input conditions such as sketches, segmentation maps, or key poses. This framework supports input bootstrapping for multiple models simultaneously, providing greater control over the generation process. For the Stable Doodle use case, T2I-Adapter provides auxiliary guidance, enabling a pre-trained text-to-image model (SDXL) to understand the outline of a sketch and generate an image based on the cues and outlines defined by the model.

The T2I-Adapter network consists of approximately 77 million parameters. It provides additional guidance for pretrained text-to-image (SDXL) models while keeping the original large text-to-image models unchanged.

How to Get and Use a Stable Doodle

The Stable Doodle is currently available on the Clipdrop by Stability AI website and app (iOS and Google Play). Users can start trialing the tool without logging in, but there is a daily usage limit. The tool's user interface is designed to be intuitive and easy to use: users simply create a basic sketch with the mouse, choose an art style, and hit "generate."

The Stable Doodle even allows for artistic customization, with 14 styles to choose from via the Stable Diffusion XL. Styles range from realistic (photography) to film to creative (fantasy art and origami).

Stable Doodle Values ​​and Limitations

Although the Stable Doodle demonstrates impressive capabilities, its inherent limitations need to be recognized. Stable Doodle uses an algorithm to analyze the contours of an image to produce a visually pleasing and coherent result. The final output depends on the initial drawing and description provided by the user, and the accuracy of the tool may vary due to the complexity of the scene.

Users should exercise caution when relying solely on the Stable Doodle for critical applications. As with all Clipdrop tools, users of the Stable Doodle are subject to Clipdrop's General Terms and Conditions.

Taken together, the release of Stable Doodle offers a whole new way to create for professionals and enthusiasts in all walks of life. Despite some limitations, its potential and possibilities remain immeasurable. We look forward to seeing more innovations and application cases.

6e8259105aa0e1f910d8ad7048ae52ec.gif

Guess you like

Origin blog.csdn.net/dQCFKyQDXYm3F8rB0/article/details/131798850