Is faster always better when it comes to creative results?

Is it faster, yes. But it also looks good and much cheaper and easier than actually shooting on location for this shot. Good work! 

I usually stay out of these AI-threads, but here there seems to be room for some real reflection 🙂 So here goes:

The question, ​@Kevin Stohlmeyer , is if it really is getting better. With 50% of the entire internet now AI-generated, cannibalization/inbreeding deterioration is a real concern. Realistically, quality can only go down without fresh outside input. For Adobe, a lot hinges on how Stock training data are filtered, because from what I hear the proportion is similar there.

For my own part, I avoid all generative AI. Not so much for the risk of six fingers, three legs and all kinds of other weird artifacts - but for its predictability. I've noticed that I can easily spot AI content simply in the way it never surprises me. It never makes me sit up and notice. It just drones on and on, with things I've seen and heard before. And, let's face it, that's how it came to be in the first place, right? The more cliche, the higher the likelihood of turning up. It's inherent in the paradigm.

Now, that predictability has a place in hard sciences. But in image editing? That has to be a very minor part of the finished work. I'll stay on the other side of the fence for now, Photoshop has plenty tools for anything I need to do, ever, without AI. I'll take the extra time.

That's just me.

The size limitations for full res outputs generally places AI results into a comp/ideation bucket for us. While you can do easier edits the trade off is spending more time with quality control when the results are not up to par. 

Don’t get me wrong AI generations are getting better but until I can produce a full res edit on a 5000 px image without decline in quality - it’s going to remain a “fun” feature with potential in the future. 

Whether genAI is actually faster or not very much depends on the context. And the quality that you’re after. In your example that shark isn't comped well into the environment and doesn’t recede in the background enough. It’s almost a mini-shark now, because the colours are too saturated and not blue enough, in my opinion.

GenAI is often just not good enough yet. For example:

QRT: missmayn
boss has been “up for days” trying to Ai generate a set design and i told him we can just hire a 3D artist to do it in less than a day.

he reluctantly agreed and now he’s “amazed” he can request little tweaks and they can do it without altering the entire image.

cost: $300 🤦‍♀️

...and just imagine the cost in terms of environment, energy, and money.