State of the art in AI image generation as we go into 2026.

Igor Petrovic

What is interesting for me while interacting with ChatGPT on various topics is its pharisee position on general ethics, when advising one how to deal with real-life problems. While being pretty much workaroundish about his moves  

I do not complain about it, just notice how it kind of resembles a widespread parenting model when some of us were kids 30-50 years ago (do not do what I do, but do as I tell you).

Chris Lozos

AI depends on the "make me one of those but a bit more like one of these" theory. What I mean is, there has to be something like it already done by someone that can be accessed and manipulated. There is no "original thought" happening. What is the point? To me, AI is only of value if it can do the drudgery part of the job, not the creative part.

John Hudson

I have experimented with giving ChatGPT image prompts based on a mental idea I have for a new typeface. I have tried to describe the idea in terms of both structure and concept, providing iterative suggestions to see if I can nudge the results anywhere even in the same ballpark as what I have in my mind. ChatGPT fails completely at this kind of thing for exactly the reason that Chris identifies: it is unable to make anything without reference to something that already exists. My prompts are all interpreted in terms of existing styles and banal associations, so the results constantly veer towards irrelevant pastiche.

It is also apallingly literal. Today, I gave the new ChatGPT image tool this prompt:

A wide image of uppercase letters, A-N on one line and M-Z on a second line, in a style inspired by the spirit of Palladian architecture.


  

Mark Simonson

Instead of artificial intelligence we get artificial smart ass. 

Craig Eliason

Chris Lozos said:

To me, AI is only of value if it can do the drudgery part of the job, not the creative part.

It feels to me like it may be precisely the opposite: AI spews plausible-but-problem-filled output, and it’s up to the humans to clean it up.

Ermin Međedović

LLMs produce output that appears plausible but contains underlying issues. Yes. Identifying and correcting those issues is left to humans. Yes. 

That, however, assumes the user can recognize what’s wrong in the first place. And they can't. For most users, “good enough” is simply enough.

Chris Lozos

Type design has always been a place for people with design talent, skills,  and perseverance enough to carry out a long task. AI seems to be targeted towards someone who wants to circumvent all of that and get instant gratification, even if the product is less than desirable. I assume the AI developers are trying to sell their software merely to make money, not to make good type.

Thomas Phinney

I think a lot of the best of what we are seeing is not “AI generates images for you out of nothing,” but rather “AI riffs on something you give it, in major productive ways.” Not quite as dramatic, but maybe more useful.

Also, actual software coding is where more of the action has been taking place lately—the big AI companies putting more effort into it.

I still expect to see AI-generated fonts (or even extensions to fonts) Real Soon Now. This year for sure.