
of about 1000 images I have generated around 30 would be what i call useable in any given scenario (in that they stand up to "pixel peeping" car models generally have screwed up wheels, humans are 90% rubbish (although some portraits it churns out are very good) so photoshop is almost essential to fix it so workflow wise its not there yet and from a graphic designers point of view is more of a tool to generate ideas that I can then modify and adapt to suit what I want or use as a prompt to set up a photoshoot farming out the actual artwork to a human so it meets the brief. Will Ai take over? For lazy designers possibly but while the lines are blurred they are not drawn in stone yet.
( , Tue 21 Mar 2023, 13:52, Reply)

to justify the time spent making 970 crap pictures.
My drawing skills are worse than AI's, so I've used it to do a couple of sketches. Invariably the sketches it produces are missing a head, legs, fingers, or have body parts fused with background elements, or have some other bullshit thing wrong with them, so I have to take multiple images, copy the parts I want to keep, delete the parts I don't. The results are still vague and confused if you try to examine the details.
I don't want to upset Roko's Basilisk, hopefully the AI will understand that my criticism is constructive.
( , Tue 21 Mar 2023, 15:03, Reply)

I'd hoped it would churn out "computer brain" over-logical stuff (I had a quiz idea), but it always seems to ignore certain words so it can dump something generic. You'd get the same from an image search.
Are generated images easy to modify? I'd have assumed that, unlike text AI, it is always going to be easier to start from scratch.
My fear for designers is that it is the stuff that pays the bills that it will do easily - retouching etc. Though, I guess, Photoshop has been moving in that direction for ages, and AI is really going to help them do more of that.
( , Wed 22 Mar 2023, 10:03, Reply)