A Brief Review of AI Art Generators

I’ve been dabbling in AI Art Generators via NightCafe for about a year now. While it is fun to play with, it it more of an entertainment thing than a serious tool at the current state of development. Sure there are a lot of people that can man-handle these things enough to get usable results. Sometimes you get lucky and get something decent out of these AI bots on the first go. Most often, however, you do not.

Personally I don’t think a useful business tool is one that takes you hours-upon-hours of tweaking, playing, and adjusting to get a result. Sure it can be less expensive that hiring a graphics professional, but with services like Fiverr these days just how much less expensive. I’ve also never wasted hours getting what I want out of a real human hired at sources like Fiverr. If they don’t get exactly what I want on the first go-around within 24 hours it is usually damn near perfect after a “do this instead of that” request that comes back less than 48 hours later. The best part is those requests to “Fiverr Humans” takes all of 20 minutes of my time.

In contrast, I’ve yet to see anyone that tries to use AI generated art or graphics not get sucked into the time wormhole and lose an hour, or multiple hours, out of their day. Instead they waste time trying to get things right. Often wasting 20 or 30 minutes to throw away a bad start that got increasingly worse, only to start over.

AI Art Generators Are A Hobby

To me, that makes AI Art generators a hobby, not a serious business tool. Not too mention that no matter how much you play with these things, the art all comes out with the same “vibe”. The only difference is the style that you coached the AI to start with, but if you pick the same engine you almost always get the same “feel” from the resulting graphics.

Case In Point

In writing my AWS Amplify articles, I decided to check in on my AI bots for a featured image. The article is on using Amplify Studio, a bot of a different nature in my opinion, to create data tables. So I asked the AI to create an image of a bot creating a table. After 45 minutes of “evolving” my creating it was still not correct. No matter how much I try I cannot get the AI bot to follow simple instructions that a human creator would have understood on the first go.

And to that point, a human would very likely have NEVER create the initial image of a bot standing IN the middle of a table as if it was part of some horrific Star Trek Teleporter accident. Humans, unlike Art AI algorithms have something called life experience. They incorporate that into their creations and rarely, unless as a deliberate act, create something with humans (or human-like android bots) bifurcated by a dining table.

Don’t get me wrong, I think this technology is fascinating. The AI generator created a much better original art piece than I could create after a year of trying. Maybe I can do a famous cartoon drawing of mine, but it wouldn’t be nearly as cool in a photo-realistic style.

That said, even though the starting image was close, further prompts did nothing more than confuse the AI Art bot even more. Move the bot backwards behind the table, just swaps out the bot with another bot style and drops the thing it was building from in front of it. Ok, let’s simplify the request “remove the foreground table” instead removes the bot, leaves the table, and replaces the both with some glass sculpture and a wine bottle?

I guess the wine bottle is a consolation gift for the human making the requests, and the bot has been recycled into glass artwork for screwing up the task so badly.

As I said earlier, great entertainment. Not really a business tool. Not yet.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.