Shulou(Shulou.com)11/24 Report--

Although CEO has been mired in all kinds of controversy before, it still does not prevent Stability AI from appearing in time magazine. Recently, the company also released the XL version of Stable Diffusion, 3.5 billion + 6.6 billion double model, equipped with the largest OpenCLIP, so that the quality of AI pictures has a new leap.

Stable Diffusion has been upgraded again!

Recently, Stability AI released the latest version of Stable Diffusion XL 0.9 (SDXL 0.9).

Compared with previous models, this wave of updates has made a qualitative leap in terms of image and composition details.

Especially in terms of parameters, this SDXL0.9 has an integrated pipeline of 3.5 billion parameter base model and 6.6 billion parameter model. By contrast, the Beta beta uses only a single 3.1 billion-parameter model.

To produce more realistic images, as well as greater depth and higher resolution (1024x1024), SDXL 0.9 uses two CLIP models, including the largest OpenCLIP model to date (OpenCLIP ViT-G / 14).

Not only that, SDXL 0.9 runs on consumer graphics cards. Only Win10/11 or Linux operating systems, 16GB memory, and Nvidia RTX 20 series graphics cards with more than 8GB video memory are required.

What is the difference between SDXL 0.9 and SDXL Beta?

Let's look at the picture and talk.

Let's take a look at the new version of SDXL 0.9. what's the difference in the details of the picture?

Prompt: aliens stroll in Las Vegas

SDXL Beta version

SDXL 0.9Prompt: a wolf in Yosemite National Park

Negative prompt:3d rendering, smooth, plastic, blur, grainy, low resolution, animation, oversaturation

SDXL Beta version

SDXL 0.9Prompt: coffee by hand

Negative prompt:3d rendering, smooth, plastic, blur, grainy, low resolution, anime

SDXL Beta version

SDXL 0.9 officially said that the SDXL series will also provide a range of features that go beyond the basic text prompt.

Includes image-to-image prompt (input an image to obtain changes in the image), inner painting (missing part of the reconstructed image), and outer painting (building a seamless extension of an existing image).

SDXL 0.9 runs on two CLIP models, including one of the largest OpenCLIP models trained to date (OpenCLIP ViT-G / 14), which enhances 0.9's processing power and the ability to create realistic images with greater depth and higher 1024x1024 resolution.

The SDXL team will soon release a research blog that describes the specifications and testing of the model in more detail.

The most influential company in Time was recently named by time magazine as one of the top 100 most influential companies.

As for Stability AI, this is what time magazine says--

If you can describe it in words, Stability AI can convert it into a picture.

Stable Diffusion, a free open source text-to-image generator that the company helped train, changed the world's understanding of the potential of AI when it was launched in August.

However, Stability AI was quickly caught up in disputes, including how to train the tools and copyright lawsuits to obtain data from the Internet.

Still, the company said the model was behind four of the App Store TOP 10 apps in the month after Stable Diffusion 2.0 was launched.

And the company CEO Emad Mostaque is also reported to often exaggerate the company's performance, after claiming that Stability AI's "real open source" paves the way for a "breakthrough."

SDXL Beta, in fact, the Beta version of SDXL has not been released for long, so the iteration of the version of mapping is really changing with each passing day.

At the time, StabilityAI said that Stable Diffusion XL was not the name of the final release, nor was it v3, because the architecture of SD-XL was very similar to the model architecture of the SD-v2 series.

The following sample pictures officially released by SD-XL show that the quality of the image is already very good.

SD-XL 's improvements over previous versions are as follows:

High quality images can be generated using a short descriptive prompt

Images that are more suitable for prompt can be generated.

The human body structure in the image is more reasonable.

Compared with v2.1 and v1.5 (to a lesser extent), the images generated by SD-XL are more in line with public aesthetics.

Negative prompts (negative prompt) are optional

The resulting portrait is more realistic.

The text in the image is clearer

Clearly readable text does not have the ability to generate readable text in pictures in v1 series and v2.1 versions of the Stable Diffusion model.

Although the text information generated by SD-XL is not always accurate, it has been greatly improved.

A young woman holding a sign with "Stable Diffusion" on it, bright hair, sitting outside the restaurant, brown eyes, wearing a skirt and better side lights. Stable Diffusion has always had a lot of problems in generating human anatomy, and more legs and fewer arms are all too common.

For example, when SD-v1.5 generates images of yoga, there are often distorted human bodies.

Although the image generated by SD-XL is not perfect, it has made remarkable progress in human posture.

More aesthetic (more aesthetic) such as the same room as the theme, SD-XL can produce more symmetrical, better visual effects of the photo.

SD-XL has also made significant improvements in portrait photos.

Reference material for a woman's photo:

Https://stability.ai/blog/sdxl-09-stable-diffusion

This article comes from the official account of Wechat: Xin Zhiyuan (ID:AI_era)

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.