I’m not saying that I’m completely knowledgeable on image making AI, but I have tested most of the commonly popular software, and some of the not so common. The trouble with reviews is that the software generally doesn’t get worse, it tends to correct itself and gets better.
For now, the best part of portrait AI is the backgrounds, the clothes, the hair, the skin tones, and the general placement of the facial features. Artistic styles can be fun to look at, but twenty anime portraits seems a bit much.
The spookiest feature is the ability to turn the head to a 3/4 shot from a full on frontal, or the ability to automatically animate the image.
Right now the easiest “tell” for an AI generated image is the pupil of the eye. The correct size, the correct highlight from eye to eye, and the general construction of the iris. The very last trial I made had nearly corrected this flaw, so it probably will disappear in the next few versions.
The text based AI is already scary, the image based AI will change forever our ability to trust photographic images.
Share this:
About johndiestler
Retired community college professor of graphic design, multimedia and photography, and chair of the fine arts and media department.
lomekian – London – Matthew Wade , or @lomekian on Twitter. Arsenal, Middlesex Cricket Club and Detroit Red Wings Fan. Professional actor, writer and voice-over artist. Former employment advisor and project manager. See my websites http://edgeofthearea.wordpress.com/ & www.matthewwade.co.uk
johndiestler – Lafayette, California – Retired community college professor of graphic design, multimedia and photography, and chair of the fine arts and media department.
AI Review
I’m not saying that I’m completely knowledgeable on image making AI, but I have tested most of the commonly popular software, and some of the not so common. The trouble with reviews is that the software generally doesn’t get worse, it tends to correct itself and gets better.
For now, the best part of portrait AI is the backgrounds, the clothes, the hair, the skin tones, and the general placement of the facial features. Artistic styles can be fun to look at, but twenty anime portraits seems a bit much.
The spookiest feature is the ability to turn the head to a 3/4 shot from a full on frontal, or the ability to automatically animate the image.
Right now the easiest “tell” for an AI generated image is the pupil of the eye. The correct size, the correct highlight from eye to eye, and the general construction of the iris. The very last trial I made had nearly corrected this flaw, so it probably will disappear in the next few versions.
The text based AI is already scary, the image based AI will change forever our ability to trust photographic images.
Share this:
About johndiestler
Retired community college professor of graphic design, multimedia and photography, and chair of the fine arts and media department.