Hey, Amy! I just read about the new Stable Diffusion 3.5 models. What’s that all about?

Oh, cool! Stable Diffusion is a set of AI models that create images based on text. The new version, 3.5, promises to make images that are more diverse and customizable.

Diverse? Like, how exactly?

Well, it means the models can create images showing different types of people, like various skin tones and features, without needing a lot of instructions. It’s more inclusive this way.

Oh, that’s interesting! Does that mean they fixed the problems from before? I remember people complaining about weird images.

Good point! Stability says they’ve made improvements, but they admit there could still be some issues with how the model understands prompts. They’re trying to make the images better while keeping different styles possible.

I see. So, what’s new in this version?

There are three models in the new series. The biggest one, called 3.5 Large, is really powerful and can create high-resolution images. Then there’s a faster version called 3.5 Large Turbo, but it’s a bit lower in quality. And, they’ll release another one for phones and laptops soon, called 3.5 Medium.

Wow, so it’s like they have different sizes for different needs! But, can anyone use these models?

Yes, kind of. They’re free for personal and research use. Small businesses can even use them commercially if they make less than a million dollars. But bigger companies have to get a license.

Makes sense. I heard there were some issues with copyright too. How do they handle that now?

They still train on public data, which can include copyrighted material. Stability allows artists to remove their images from the training set, though, and they say they’re following fair-use rules. But, if someone sues, it’s up to the user to defend themselves.

Oh, that sounds risky! Are there any new safety features, especially with elections coming up?

Good question. Stability says they’re trying to stop misuse, like creating fake information. But they didn’t share much detail about how they do it.

I hope it works. It sounds like they’ve improved a lot, but there’s still more to figure out.

Exactly. AI art is growing fast, and models like these will keep changing. It’s exciting to see where they go next!