>Photoshop's generative AI is substantially worse than what I can get from models on Tensorart/CivitAI + an upscaler.

But with Photoshops, I can generate things to be seamlessly placed into my existing scene, with it understanding things like placing the wheels of the car it just inserted on the ground, with a shadow underneath and such, again seamlessly integrated with my existing image.

It's very much not just about generating an entirely new AI image.

Can I do that with Tensorart/CivitAI?

Isn't that just inpanting? Stable Diffusion has had it for a while.

Eg: https://www.youtube.com/watch?v=No1_sq-i_5U

You can do that with any part of the image, this video shows a face sample.

It might be.

https://youtu.be/Sp6K3qpVFO0?t=95

But at the very least, it seems like it's easier to work with in Photoshop.

Worse or not, a meaningful majority of users will never even understand the alternate workflow you just described.

Then they will be outproduced. I, a graphic design noob, managed to create a bunch of social media posts for an upcoming project that included a completely custom 3D character in precisely the poses I wanted, all within an hour. Without gen AI, I would have likely spent hours digging through stock photos, modifying lighting and colors, and still not achieved what I wanted to.

Not using the latest and greatest tools in your profession isn't really a flex.

Maybe they are hampered by "doing the right thing" and only training on Adobe Stock content?

Firefly (Adobe's generative AI) works well if you're already stuck in a Photoshop-heavy workflow, for quick successive generations right in the canvas if you're editing or extending an existing image. Better than Stable Diffusion Photoshop plugins I've tried.

I think what you mean to say is Stable Diffusion. CivitAI is primarily a host for SD models to download and run locally.

You can run many of the models directly through CivitAI. You get lower quality images, but you can upscale them through an upscaler.

Just go to any model and click "Create". You can even use examples from the model to prefill the prompt.

For example, you can scroll down and click "Start Creating" on this page: https://civitai.com/images/3908138

CivitAI’s web tools are great for someone getting their feet wet but it presents an extremely stripped down options UI relative what’s available in Draw Things on iOS/macOS or SHARK on Windows.

True, but that's precisely my point: even with this stripped down version, I was able to hack together exactly what I wanted for an upcoming project in far less time. I got a model in the exact hairstyle, pose, and costume I wanted. The upscaler did a pretty good job of making it production ready.

This stuff is wildly productive for noobs who just want to get stuff out. Maybe its not something you'd see in an Apple ad, but its definitely something you'll see in a small business ad.

I couldn't care less about AI in Photoshop but I'd love AI in Illustrator to generate vector stuff.

Text to Vector appears to be a beta feature in Illustrator currently: https://helpx.adobe.com/illustrator/using/text-to-vector-gra...

AI sucks pretty hard at generating Vectors. The only option for now is to generate something in an illustration style and have it vectorized.

There are a couple of AI vectors services out there but the pricing is just silly.

https://vectorart.ai/

https://www.kittl.com/feature/ai-text-to-vector

Adobe's vision is fully supporting the content lifecycle, not only the creative aspect. In this sense, Figma made far more sense for Experience Cloud rather than what it gives to Creative Cloud.