All presented images are generated with the WAI-SHUFFLE-NOOB
Description
FAQ
Comments (7)
Excellent 😍
Why don't you include all the prompts for the outfits? Are they burned into the character maybe?
by default she goes in her anime clothes, since the dataset has many images of her in this outfit, the dataset also has examples of how she looks in a school uniform, if you put the tag school uniform, then there will be a school uniform like in the anime. It makes no sense to describe all the outfits since it can be anything.
@fakerrak I've heard tagging/prompting the outfits helps make it more flexible. Thanks for the reply and sharing the Lora.
@fakerrak I've heard tagging the outfits can make them more flexible and make it easier to create them in different outfits. Thanks for sharing the Lora and replying.
@Lancelot69lol Yes, ideally, all parts of clothing should be described in the caption of dataset images, but in practice this is difficult since clothing can be quite complex
@fakerrak Sorry for the double reply, my first message didn't show up so I wrote it again but it's here now (Civati being broken lol) I agree with clothing being tough to properly tag but you can get a lot of info for them using an autotagger or checking Danbooru.
It's a lot more work no doubt but the results are often worth it. Not saying you need to do it, do what works for you and again thank you for sharing your Lora.