/vt/ - Virtual Youtubers

>>60673147
>I'm not sure they'll do anything but I can at least try it, I guess
Ok well it definitely did something, though I'd have a hard time describing exactly what. It seems a little less wild with colour, like it's more consistent/gradual instead of just sudden bursts of very bright completely different colours
https://files.catbox.moe/z92vsj.png
https://files.catbox.moe/1ebr05.png

AI still has lots of issues with weapons though
https://files.catbox.moe/g5unjp.png
https://files.catbox.moe/50o6gb.png
https://files.catbox.moe/tmyoa3.png

I think that's enough prompting for me for now. nice request idea, reminded me how fun it can be prompting wild scenery >>60673697
>I don't understand much of how AI works but those prompts are from AI images I found on the internet that fits
I'm not an expert either but I think most anime-style models are derivatives or at least have heavy influence from the NAI leak, which had training data mostly tagged with danbooru-style tags and so when you prompt with it you generally try to use prompts built by stringing tags together. Not to say it can't recognize general language prompts to some extent but that sort of prompting is more appropriate for something like dall-e (which unfortunately I don't think generates anime style that well. I think.). At least, that's my understanding.
Your initial request, just copied word for word, would probably make a decent dall-e prompt. Well, you'd probably replace "ayame" with "a girl" or "a woman" or something.
>Do you think the AI can't handle a magic swordsman?
It's just that holding things correctly tends to be a bit of a struggle