Stable Diffusion XL is here — What’s new?
Today, Emad Mostaque, Founder and CEO of Stability AI, announced in a Tweet that Stable Diffusion XL testing is now available for public testing.
Here are some things you need to know:
- “XL” is not the official name of the new AI model. It’s going to change once Stability AI makes an official announcement
- The quality of images is improved compared to the previous version
- Image generation is a lot faster compared to the previous version
Example Images
Let’s take a closer look at the differences in the results between the new and old AI models.
Prompt: Luxury sports car with aerodynamic curves, shot in a high contrast, high key lighting with shallow depth of field, exotic, detailed, sporty, studio lighting, HQ, 4k
Prompt: Bowl of steaming hot ramen with a perfect egg in the center, surrounded by thin slices of meat, green onions, and nori, with a flavorful broth and perfect noodles, high detail, focused on texture and steam
Prompt: Enchanting waterfall in a lush jungle, surrounded by exotic plants and wildlife, tranquil, serene, high detail, tropical landscape
In all honesty, I don’t really see any drastic improvement with the new AI model. Yes, in some cases, it produces better-looking images, but perhaps it needs a few more tweaks in the settings to get it to produce jaw-dropping results.
In terms of speed, I noticed an improvement, but not that much.
How about you? Do you like the new Stable Diffusion XL?
How To Get Access?
Stable Diffusion XL (beta) is accessible on the new Dreamstudio website.
You can also access the model on the old website.
How Much Is It?
You will be billed based on how many generation credits you use.
$10 = 1000 generation credits.
This is how much credit you will use, depending on the image configuration you set.
If you leave all settings at default, 1000 credits are approximately 5000 images.
Final Thoughts
Overall, I am not that impressed with the results of the new preview model. MidJourney, Leonardo AI, and Microsoft Image Generator still have better-looking results.
Although Stable Diffusion XL may not seem like a significant improvement over the previous AI model, it is still a step forward, and there is certainly room for further improvements.
I am excited to see what the company will reveal with its official announcement, and I have no doubt that the final product will be even better than the preview model.
Stay up to date with the latest news and updates in the creative AI space — follow the Generative AI publication.
Please support my work on Medium and get unlimited access by becoming a member using my referral link here. Have a nice day!
