ghost
Sign In

Dynamic Waifu

42
742
9
Updated: Jul 31, 2024
toolanimesexywomancutegirls
Type
Wildcards
Stats
553
Reviews
Published
Jul 31, 2024
Base Model
SDXL 1.0
Hash
AutoV2
F1B56CE2D6

This set of wildcards was created to aid in the generation of beautiful women beautiful. The initial release focuses on solo women, but I’m open to creating more wildcards for other subjects, groups, etc., if there is sufficient interest. Feel free to leave a comment to let me know what you may want ot see in the future. If you couldn’t tell from the name, this wildcard set is intended to be used primarily with anime or semi-realistic models or models with a fantasy or sci-fi theme. Because it's based on Danbooru tags, it works very well with Pony models. It’ll also work perfectly fine with photo-realistic models but expect the anime visual characteristics (non-traditional eye and hair colors, for example) to push your girls into the uncanny valley. These wildcards were specifically designed for and tested with Deep Dark Hentai Mix version 6.1. I’m partial to the model, but I promise you won’t be disappointed in the results.

Usage

The wildcard files are in a nested YAML format consisting of arrays, strings, and weighted dynamic lists. If that sounds complicated, don’t worry, you don’t need to know any of that to use the wildcards. For the basics, here’s what you need to do:

  • To use the wildcards, simply unzip the downloaded file and drop the waifu folder and all its contents into your wildcards directory. You can use the instructions at https://github.com/adieyal/sd-dynamic-prompts to make sure you have the required dynamic prompts extension (or nodes) installed and that the wildcards directory is available and set up.

  • There is already a text file in the directory that contains one wildcard to rule them all. As long as you placed the waifu folder in the correct location, simply prompt "__waifu/waifus__" to generate images.

  • For more advanced prompting, you can pick and choose from these wildcards, which make up the elements in the simple wildcard: “__waifu/body__”, “__waifu/skin__”, “__waifu/eyes__”, “__waifu/hair__”, “__waifu/nail-polish__”, “__waifu/lipstick__”, “__waifu/jewelry__”, “__waifu/makeup__”, “__waifu/happy-expressions__”,”__waifu/poses__”

  • To learn more about what is included, in the section below.

If you are an advanced wildcard user, you can check out the files and the wildcards listed under the wildcards manager. You’ll see the wildcard files actually contain a pretty large number of wildcards you can use, including three large arrays for clothing, facial expressions, and jewelry items. Any can be used to prompt only select elements, but the top-level wildcards listed above are really all you need if you don’t want to get these too complicated.

What’s Included in the base version

If you want more variety but still want to keep it simple, the top-level wildcards can be used on their own. Here’s what each is supposed to do.

  • waifu/body: made up of three sets of descriptions for body build, figure descriptions, and breast size/descriptions.

  • waifu/skin: includes skin details like freckles, moles, tanlines, skin tone, and tattoos.

  • waifu/eyes: anime eye colors in a weighted list. Standard colors are more likely (brown, blue, green), but other colors will popup as well. You can always set their probability to 0 in the wildcard file if you don’t want a certain color to appear.

  • waifu/hair: will prompt hair colors, hairstyles, and hair accessories. Like eye color, hair color can be controlled by adjusting probabilities.

  • waifu/jewelry: will pick 0 to 5 items from rings, bracelets, earrings, necklaces, or miscellaneous jewelry (e.g., toe rings, eyebrow piercing, etc.).

  • waifu/makeup: will pick 0 to 5 items from eyeliners, eyeshadow, mascara, lipstick, or nail polish. Instead of using this wildcard set, you can prompt for nail polish and lipstick by themselves.

  • waifu/nail-polish: for fingernail polish. Fingernail color and some descriptions (e.g., long nails, sharp nails, etc.) are included.

  • waifu/lipstick: simple list of lipstick colors. Red and pink are most likely to show up.

  • waifu/happy-expressions: a select array of happy, positive expressions. There is a large expressions array, so others can be picked as well. Use waifu/expressions/* to pick random expressions from the entire array.

  • waifu/poses: the most complex part of the wildcard set. These include sitting, lying, standing, and miscellaneous other poses. These poses have set objects and location variables (so a woman will lay on a towel at the beach, on a blanket at the park, or on a couch in a living room) and are also tied to clothing sets appropriate for the location (i.e., swimwear for the beach). Most of the time, these will work properly, but occasionally, an odd generation will pop up.

About the base SFW version

The base version using the "__waifu/waifus__" wildcard will not create prompts that specifically generate images that contain nudity or explicitly sexual actions. Clothing does include undergarments and swimsuits, and some postures could be interpreted as suggestive in nature. Keep in mind that generation is controlled by the model and some models will generate nsfw content no matter what you use to prompt. If you want to allow nsfw prompts, you’ll need to go into the files and switch a few things around to enable it or use the general wildcards you can find in the wildcard manager. Either way, I suggest adding negative prompt elements to discourage any generation of minors. For Deep Dark Hentai Mix version 6.1, add “loli, child” to the negative prompt.

Generation Notes 

Prompt Formatting: Due to the complex nature of the YAML coding, there may be a few extra spaces or commas in the prompt. Not to worry, it shouldn’t effect generation.

Color Influence: Stable Diffusion can get confused with a lot of colored elements in the prompt. These wildcards do use some colors (hair, eyes, makeup, etc.), so you may see these elements carrying over to match other parts of the image (clothing for example).  It isn’t intended, but does tend to improve the image composition by coordinating elements.