AI is better than it was, but it's still no panacea. If you're gonna try AI for placeholder people I'd advise using StarryAI or Stable Diffusion - and try to get it to give you images without any hands... like tell it you want their hands in their pockets, or wearing mittens, or something like that because it's rare for an AI generated person to have the correct number of fingers!
There may be other AI platforms that do a decent job, but those are the only two that I've had success with. My original attempts were using DALL-E, and while people generated by DALL-E have improved, they still tend to look a bit like something from a bad horror movie! However, DALL-E's outpainting feature can be useful for doing touch ups and/or for expanding the backgrounds if you get a decent picture from a different platform with a mangled hand, or someone's head cut off, or not centered or something.
I've been able to get some useable stuff - but it does take work both in terms of learning how to give it the right cues, and also editing them by hand afterwards - I've yet to get anything AI generated that's useable as is - all of these have several hours of post-generation work put into them. Sort of a pain, but I guess it's good in that means nobody else is gonna be able to get it to dump out the exact same image. I wonder what that means in terms of copyright... hmmm... but I digress.
Anyhow, here are a few examples:
Placeholder image generated by Stable Diffusion. The original image was not centered and didn't have enough background, plus there were a few extra arms in there, so it was edited extensively with DALL-E and also by hand, and enlarged using Topaz
AI is better than it was, but it's still no panacea. If you're gonna try AI for placeholder people I'd advise using StarryAI or Stable Diffusion - and try to get it to give you images without any hands... like tell it you want their hands in their pockets, or wearing mittens, or something like that because it's rare for an AI generated person to have the correct number of fingers!
There may be other AI platforms that do a decent job, but those are the only two that I've had success with. My original attempts were using DALL-E, and while people generated by DALL-E have improved, they still tend to look a bit like something from a bad horror movie! However, DALL-E's outpainting feature can be useful for doing touch ups and/or for expanding the backgrounds if you get a decent picture from a different platform with a mangled hand, or someone's head cut off, or not centered or something.
I've been able to get some useable stuff - but it does take work both in terms of learning how to give it the right cues, and also editing them by hand afterwards - I've yet to get anything AI generated that's useable as is - all of these have several hours of post-generation work put into them. Sort of a pain, but I guess it's good in that means nobody else is gonna be able to get it to dump out the exact same image. I wonder what that means in terms of copyright... hmmm... but I digress.
Anyhow, here are a few examples:
Placeholder image generated by Stable Diffusion. The original image was not centered and didn't have enough background, plus there were a few extra arms in there, so it was edited extensively with DALL-E and also by hand, and enlarged using Topaz
