Stable Diffusion 3 Medium: The Successor to SDXL

M. Yekta Güngör

The company behind SDXL announced their latest model a month ago: Stable Diffusion 3 Medium. Stability AI describes it as their “most advance text-to-image open model”. It comprehends your prompts significantly better and is able to incorporate text into images. Let’s take a look at what it offers.

Text in Images

One of the big shortcomings of SDXL was that it couldn’t incorporate text into images. Stable Diffusion 3 Medium can do that. However usually it takes a couple of tries to get the spelling right.

Text Example 1

Prompt: a cute magical creature in front of a big purple neon sign that says “This is nice!”, cinematic, a city in the background.

a cute magical creature in front of a big purple neon sign that says "This is nice!", cinematic, a city in the background

Text Example 2

Prompt: balloons shaped as letters to make the word “autumn”, a high end home in the background, cinematic, autumn colors, old Leica photo.

balloons shaped as letters to make the word "autumn", a high end home in the background, cinematic, autumn colors, old Leica photo

Better Prompt Understanding

Stable Diffusion 3 Medium has a much better understanding of your prompts compared to SDXL. This comes in handy when you are trying to generate very specific images. For example if your images requires a red bottle on the left and a blue bottle on the right, it can do that.

Example 1

Prompt: ancient magical translucent bottles, left one has a red liquid in it, middle one has a green liquid in it, and the right one has a blue liquid in it, an ancient room in the background, Unreal Engine render.

3 ancient magical translucent bottles, left one has a red liquid in it, middle one has a green liquid in it, and the right one has a blue liquid in it, an ancient room in the background, Unreal Engine render

Example 2

Prompt: 3 cute medieval cats wearing dresses, left one has a purple dress, the middle one has a yellow dress and the right one has a teal dress, cinematic, dark tones, bokeh.

3 cute medieval cats wearing dresses, left one has a purple dress, the middle one has a yellow dress and the right one has a teal dress, cinematic, dark tones, bokeh

A Not So Permissive License

Lately, some open-source models have been switching to non-permissive, research only licenses. This means that you can’t use those models for commercial purposes without purchasing a license. Stable Diffusion 3 Medium was released with such a license, having heavy restrictions on what you can do with it commercially. Just a couple of days ago Stability AI modified the license and made it significantly more permissive.

Under the new license, Stability AI says that you don’t need to pay them anything for commercial usage as long as your annual revenue doesn’t exceed 1 million USD.

Stable Diffusion 3 License Update

Conclusion

Considering it can understand your prompt better than our default model, it can do text in images, it is relatively fast, and its not so permissive license doesn’t affect Stablecog, we are making it the default model on Stablecog going forward.

Give It A Try

You can try Stable Diffusion 3 on Stablecog right now! Just click the button below and see what it can do.

Try Stable Diffusion 3

Thanks for reading!

Back to Blog

Read More

Create amazing art in seconds with AI.

Start Generating

© 2023 Stablecog, Inc.