Hi Just a tip For the final version
I recommend using Chinese ,Japanese ,Korean art from Twitter For training this model their art is so good ✨🤞 ;D:Be careful some of them might be Ai
Danbooru includes tons of twitter art.
If you want to see chinese/korean artists on danbooru, you can search for "chinese commentary" (I dunno if Anima was trained with meta tags like this).
The stuff you posted falls pretty neatly into the "no humans" category as well.
For japanese artists specifically, danbooru assumes that "commentary" is japanese by default, so you can exclude english and chinese when searching for commentary.
Danbooru includes tons of twitter art.
If you want to see chinese/korean artists on danbooru, you can search for "chinese commentary" (I dunno if Anima was trained with meta tags like this).
The stuff you posted falls pretty neatly into the "no humans" category as well.
For japanese artists specifically, danbooru assumes that "commentary" is japanese by default, so you can exclude english and chinese when searching for commentary.
Twitter has more than that website
Short answer: No thanks.
Slightly longer answer: Unfortunately, when collecting datasets, there's no compelling reason to spend the developer's time manually downloading, organizing, and tagging images directly from X. The time spent browsing X for decent images, gathering, sorting, and tagging them would be better spent using an already-organized Danbooru dataset. In the case of Anima, some tagging adjustments like adding natural language are also needed, which adds uncertainty. On top of that, it would also require checking whether images pulled from X are already included in the dataset and filter out duplicates. I hope that's a sufficient explanation.
I'm not dismissing your enthusiasm, but to be honest in a way that's helpful to everyone including yourself, this kind of suggestion is like a little kid running into a restaurant kitchen and asking the chef to mix their favorite snacks into the dish.
There's certainly room to interpret it charitably, but the reality is that more often than not, that's not how it lands.
i know all that
I recommend using Chinese ,Japanese ,Korean art from Twitter For training this model their art is so good ✨🤞 ;D:Be careful some of them might be Ai
AI学習禁止(AI Learning Prohibited)
Legally meaningless, but X has a lot of this kind of art 😂