Stable Diffusion Clay Style Image Generation Guide

This article is authorized to be reprinted from: 58UXD (ID: i58UXD)

A new wave of creativity is sweeping across social platforms, which is the craze of “Clay AI.” This ugly yet adorable style has become a hot topic among internet users. Whether on Xiaohongshu or other social platforms, tags like #Clay and #MyClayWorld dominate this novel creative style. Works in this clay style have overturned the traditional definition of beauty, making the expression of art richer and more colorful. They seem to mark the beginning of an art revolution, redefining people’s perception and understanding of beauty.

Today, I will share how to use Stable Diffusion to generate images in the clay style.

You can try these three large models: ReminiClay clay style, Ugly Cute Internet Celebrity Clay Style Filter, and Internet Celebrity Clay Style Filter 1.5. They can directly generate images in the clay style.

For example, using the ReminiClay clay style large model:

Keywords:

A penguin walks on the grass, high quality, masterpiece, rich detail, Clay,

Negative words:

ng_deepnegative_v1_75t, (badhandv4:1.2), (worst quality:2), (low quality:2), (normal quality:2)

Sampling method: DPM++ SDE Karras

Number of iterations: 30

Width and height: 1024px * 1024px

You can choose realistic or 3D large models as the base model, depending on the needs of the generated image. For example: Maiqi Realistic, Mo You Artificial Human, AWPoly3D, 3D Character IP, etc.

Method for Using Text to Image Generation

The operation is very simple; just choose the large model and input the parameters below into SD to wait for the image to be generated.

Keywords:

A girl walks on the street, Clay, upper body, close-up, real, masterpiece, high quality, detail,

Negative words:

ng_deepnegative_v1_75t, (badhandv4:1.2), (worst quality:2), (low quality:2), (normal quality:2)

Sampling method: Euler a

Number of iterations: 30

Width and height: 512px * 768px

Isn’t it simple? Next, let’s take a look at the popular image-to-image generation.

Method for Using Image to Image Generation

As an example, I chose the Lora model: Clay World SD1.5. Add a reference image, which I used:

Parameters can follow my settings:

Large model: ReVAnimated_v122

Keywords: A black ceramic robot that looks like a bear, masterpiece, best quality,

Negative words: low quality, lowres, watermark, worst quality, beard,

Sampling method: Euler a

Number of iterations: 30

Width and height: 1024px * 1024px

We can get a cute mechanical bear clay figurine. Next, let’s try the effect with characters. Here, I will use this AI-generated character image as an example. The clearer the image, the better the quality of the image-to-image generation, making it closer to the clay feel.

Let’s start with a simple approach:

Large model: ReVAnimated_v122

Keywords: Clay World, best quality, masterpiece, 1 boy,

Sampling method: Euler a

Number of iterations: 30

Width and height: 1024px * 1536px

We obtain the following image:

It seems that the large color blocks and concepts are already quite close. However, there are still some adjustments needed, such as the beard. You can add the beard concept to the negative words to block it.

Large model: ReVAnimated_v122

Keywords: Clay World, best quality, masterpiece, 1 boy,

Negative words: beard

Sampling method: Euler a

Number of iterations: 30

Width and height: 1024px * 1536px

We obtain the following image:

It was found that by blocking the beard, the eyebrows were affected. Here, we add eyebrows to the positive words to strengthen the concept of eyebrows.

Large model: ReVAnimated_v122

Keywords: Clay World, best quality, masterpiece, 1 boy, eyebrow

Negative words: beard

Sampling method: Euler a

Number of iterations: 30

Width and height: 1024px * 1536px

We obtain the following image:

It seems that the eyes are also not quite right. Let’s add the concept of eyes to the positive words.

Large model: ReVAnimated_v122

Keywords: Clay World, best quality, masterpiece, 1 boy, eyebrow, eye

Negative words: beard

Sampling method: Euler a

Number of iterations: 30

Width and height: 1024px * 1536px

We obtain the following image:

It looks okay. Due to space constraints, I won’t elaborate further.

So far, we have obtained a pretty good image. On the KuaiPai AI platform, you can experience ultra-fast image generation for free, and it also provides deep customization solutions for image generation for frontline businesses. It offers a stable and quick interactive experience and image generation results for non-designers, helping various departments meet their image needs.

Next, let’s look at some cases; it’s still very fun:

This is the content of this sharing.

-END-

Article reprinted: 58UXD, copyright belongs to the original author

Original link: https://mp.weixin.qq.com/s/nkFmdaRAUJXsZkuOg1jTZQ

Copyright statement: “IXDC” will indicate the author and source of the articles it pushes unless it is truly impossible to confirm. Our public account maintains neutrality regarding the content, statements, and opinions of the articles shared. We do not guarantee the accuracy, reliability, or completeness of the content, whether express or implied, and it is for readers’ reference only. Some articles may not have been able to contact the original author when pushed. If any copyright issues arise regarding the content or works, please contact us, provide the URL of the content, and relevant proof materials. We will verify and immediately correct or delete the relevant content! Our public account assumes no responsibility and has the final interpretation of this statement.

Contact WeChat: 18802086168

Contact phone: 18802086168

Editor | Ye Xuan

Final review | Su Jing

▼



点这里，学习更多设计知识！
👇