Photorealism is well within current capabilities. Technical drawings absolutely not. Not sure what other graphical media includes.

> Not sure what other graphical media includes.

I'd want a model that can draw website designs and other UIs well. So I give it a list of things in the UI, and I get back a bunch of UI design examples with those elements.

I'm gonna hazard a guess and say well within the capabilities of a fine tuned model, but that no such fine tuned model exists and the labeled data required to generate it is not really there.

You'd have better luck with an LLM with HTML/JavaScript/CSS.

Theres a startup doing that named galileo_ai

Yeah but try getting e.g. Dall-E 3 to do photorealism, I think they've RLHF'd the crap out of it in the name of safety.

well that's what you get with closed ai.

That's why we need open AI which scoops up all the data with its specific contexts and history and transforms it into a vast incomprehensible machine for us peons to gawk at while we starve and boil to death

low quality discourse imo

That's not safety, the safety RLHF is because it tries to generate porn and people with three legs if you don't stop it.

It has the weird art style because that's what looks the most "aesthetic". And because it doesn't actually have nearly as good enough data as you'd think it does.

Sora looks like it could be better.

Photographs, digital illustrations, comic or cartoon style images, whatever graphical style you can imagine are all easy to achieve with current models (though no single model is a master of all trades). Things that look like technical drawings are as well, but don't expect them to make any sense engineering-wise unless maybe if you train a finetune specifically for that purpose.

And will the model also pretend that a certain particular race doesn't exist?