Be able to grab dataset images from a website/search engine

I have a suggestion for the onsite LoRA trainer: Be able to grab dataset images from a website/search engine (using API).

Why? Can’t we just download images then put them in your dataset?

I know the method you described is possible, but it could be effortful if you don’t use a service or tool that bulk downloads images. Some may also have low storage space, and this feature would be essential for them (so they don’t have to delete files to free up storage space or use a different device with high storage space).

What websites will we be able to grab dataset images from?

Onsite (Civitai), boorus (e.g. Danbooru, E621, etc.), search engines (e.g. Google, Bing, Yandex, etc.), even other image generation sites (e.g. Perchance, PixAI, Tensor.Art, SeaArt, etc.). Not to mention, we also need an ability to grab dataset images from Fandom wikis (both images from editing and forum mode), since most of these wikis have high-quality official art on the editing mode, and high-quality fanart on the forum mode, making Fandom wikis a great source for official art and fanart.