That’d be me. I don’t really care to “keep up” with things. What should I be using instead?

Stable Diffusion but it also has artifacts as discussed in the post

Stable diffusion finetunes and model merges. You can find most of the newer ones at CivitAI.

https://civitai.com/

SD still has the hand problem, but a lot of the newer checkpoints are getting pretty good at hands.

As far as keeping up with things, the best way to do this is actually to be in AI related discords. I'm not sure why, but people involved in AI are terrible about having a single source of info...or even just info in general. I would start by joining the LAION discord then finding others from there.

Another great way to learn about what's going on in AI is actually through threads on 4chan's /g/ board. If you're willing to look past the fact that it's 4chan you will be the first to know about lots of neat stuff.

Btw, I'm not a researcher. I'm just talking about ways to get a quick scoop on what the current AI meta is and how far it can be pushed.

edit: I just realized you don't care to keep up...lol. Oops. I'm still leaving this up because I think some people will find it useful.

You can position the body and fingers and way you want with ControlNet.

> Using Dall-e 2 for all your example images is a great self report that you don't keep up much with AI.

I think the overarching story (subtitle of the article) is what is the message: "machines can grasp small patterns but not the unifying whole."

Amateur human illustrators might also struggle to draw accurate representations of hands, but, unlike machines, can critically evaluate and KNOW whether they got it right or not...

Luckily there sits sometimes a human in front of the screen to decide if the output is good.