/vt/ - Virtual Youtubers

>>56725765
I think if I had to pick two things I really really liked from this sampling, its the CrystallineAI effect and StainedGlassAI effect from the Decoration Bundle. Both are pretty unobtrusive and useful.

>>56727130
As someone who started recently, you gotta pace yourself and set achievable goals. Go on CivitAI and look at model checkpoints (you can filter by Checkpoints). Look at the sample galleries for the kind of art style they give. Don't worry about mixing or any of that shit, just download, plug something in, and prompt. Most anime styled checkpoints are trained on Danbooru data, so they expect to be fed Danbooru tags in the prompt. You'll also probably need something called a VAE; sometimes these are baked in and sometimes they're separate files.

I use this model
>https://civitai.com/models/53761/goofball-mix
which has a baked in VAE. I'm a lazy bastard but it works for me.

For characters, you usually need a lora; there's a general purpose HLL5 linked in this thread, or you can find loras for pretty much every holo on CivitAI. Some poses/concepts need a lora to give them that extra push, so if you can't get it to do wet clothes to your satisfaction, try downloading one of those. You may also want "textual inversions", which are just fancy ways of persuading the AI to avoid certain things. These usually have names like "EasyNegative" and "bad-hands".

For extensions, start with
>a1111-sd-webui-lycoris
>a1111-sd-webui-tagcomplete
>adetailer

Anything else is gravy. Don't overcomplicate your workflow when you're still figuring out how to walk. In my two weeks of prompting, I've found LESS IS MORE. It's tempting to write a 200 token prompt to but I find its much easier to give it a few loose directions and then let it do all the work.