Introduction to Stable Diffusion

Introduction to Stable Diffusion
I previously mentioned that I participated in a beta test for a new AI painting software. The official version is about to be released, and I have received authorization to publish images. All the images in this article come from this software called “Stable Diffusion”; feel free to enjoy them.
Introduction to Stable Diffusion
All these images used the same descriptive prompt: the clouds of Yunnan by XX, oil on canvas. 云南的云,XX绘制,布面油画. Of course, for some paintings, I specified colorful clouds, which made the colors look richer. You can see that a very simple prompt can yield different styles of works based on different artists’ styles. For example, the one above immediately reminds me of the years in Shangri-La, with a very realistic style.
Introduction to Stable Diffusion
A major advantage of Stable Diffusion is its optimization for character modeling, but today is the “Clouds of Yunnan” feature, so I didn’t showcase faces or human bodies. In my personal experience, it performs better than the current Disco Diffusion but is weaker than Dall·E 2. However, the speed is indeed not slow; the average generation speed for all these images is 5-6 seconds.
Introduction to Stable Diffusion
I selected a series of artists, and Stable Diffusion does a great job of aligning with artists’ styles, feeling stronger than the two mentioned earlier. But it’s also possible that I’m biased because I like their styles, so once the official version is released, everyone can experience it themselves.
Introduction to Stable Diffusion
I am very satisfied with the artwork above, which reminds me of the regions in Yunnan where the mountains are higher than the sky, and the small valleys filled with houses. Stable Diffusion handled it very exaggeratedly, but in my opinion, it is just right, truly achieving mountains higher than the sky, and valleys like water bowls.
Introduction to Stable Diffusion
Starting from this piece, the next few are in the style of the artist Victo Ngai. I can’t help it; I really like her painting style, so I’m being a bit selfish. The piece above feels like a memory for those who travel to Yunnan; it seems like your memories in Yunnan. As a member of a minority ethnic group, I have only worn ethnic clothing once in my life, and that was the Yi people’s attire, just for a photo at the Stone Forest.
Introduction to Stable Diffusion
I really like these tones, the flat treatment of the images, and those textures. In short, I just like it; I can’t be bothered to make up reasons. I’ve seen clouds like this, cumulus clouds that look like fortresses, that you can see anytime in the summer when you look up.
Introduction to Stable Diffusion
This image reminds me of the feeling of looking at Dianchi Lake or Erhai Lake from the air. Even if it’s Dianchi, because there’s a long corridor extending into the water below. I saw skies like this when I was young, but I rarely see them as an adult. That sky color I love seems to have disappeared with the city’s expansion; I don’t know where it went.
Introduction to Stable Diffusion
Finally, I hesitated for a long time about this last piece because it feels a bit too girly, which is a bit embarrassing. But what can I do if I like it, and what can I do if it looks good on me? So I’m posting it anyway.
If you are interested in Victo Ngai, oh no, I mean Stable Diffusion, after reading this, I have provided an application link for testing at the end of the article in the “Read the Original” section. Although it’s also a futures application, you probably won’t have to wait for months like Dall·E 2; you can apply and patiently wait for an invitation.
Good night.

Creator: He Cai Tou’s Little Meat Hand

AI Algorithm Provider: Stable Diffusion

Stories by the CurbProduced by He Cai Tou

[WeChat ID]Bitsea

Personal reposting of content to Moments and group chats does not require special copyright permission.

Please believe me:

Every word I say,

is wrong

Zen Moment

Writing this feels a bit of a loss; I put a week’s worth of cover images into one article, and I feel like I’ve lost out.
Thanks again to Xiqiao, thanks to Dango233, and thank you all for wanting to play with me.

Leave a Comment