Thursday, October 1, 2026

Info 2393: AI image creation by text description

Mosquito commented on my recent AI post and even gave a sample. Thanks! Gemini doesn't work without registration, at least not in Germany. Half of the Google results use Gemini or similar and so don't work either. 

But I found some that work, including Bing, a German one and 2 others. One even made an animation but took about 10 minutes for a 5 second brief clip.

Bing makes four pics from one description. Results are quite good. I tried overalls and non-overalls subjects like general teens and/or rooms. Often the age is not correct, when I choose 14yo they look like 10-11. 

The backgrounds are probably random and a surprise and often clearly look separated and people just put over them. When I choose 80s, it's very 80s. Often too much with many stuff, posters, toys, video games, computers, shirts etc.

Today I tried "teen boys in bib overalls rear view". The output showing 3 older boys. Stlye of the bibs is the same and buckles on the rear. Bibs are not tight. One is maybe a girl. I also tried "family in bib overalls like the Waltons". 2 kids are very young. A girl wearing bibs missing the legs, or might be skirtalls.

The German site has great faces and very realistic people and room. But bibs rear view are not realistic at all. Rear views are probably more difficult as most pics showing front.  

But most AI pics can still be instantly identified as such. Maybe it takes one 2 more years for perfect pics. If I don't like the heads or haircuts, photoshopping them is another option. But I prefer real movies and series.

 

No comments: