Note: Most popular generators support Lycoris LoRAs right out of the box. You might have to put it in a separate "Lycoris" folder, but if that doesn't exist then it goes in the LoRA folder.See also:<a rel="ugc" href="https://civarchive.com/models/227128/cleavagehelperxl">CleavageHelper for SDXL</a><a rel="ugc" href="https://civarchive.com/models/649980">CleavageHelper for FLUX</a><hr />The purpose of this Lycoris is to help attain the best type of cleavage. You'll get the good results if you have the word "cleavage" in your prompt. It is not recommended to be used on anime models since this was trained on many photos on the SD 1.5 base model, but you can try if you want and see if it works well enough for you.If you have trouble getting results for women with darker skin, include "dark skin" or "very dark skin" in the prompt.All images were generated using the Vladmandic fork of AUTO1111. Since I had to manually enter in the details for the preview images, here is the missing information I could not enter:For the woman in the bikini the missing information is:Face restoration: CodeFormer, Size: 512x1024, Model hash: e9d3cedc4b, Model: realisticVisionV40_v40VAE, Version: 5eb9f38, Parser: Full parserFor the woman dressed as a Bunny:Face restoration: CodeFormer, Size: 512x512, Model hash: 59ffe2243a, Model: chilloutmix_NiPrunedFp16Fix, Denoising strength: 0.7, Version: 5eb9f38, Parser: Full parser, Hires upscale: 2, Hires upscaler: LatentFor the golden...armor lady:Face restoration: CodeFormer | Size: 512x768 | Model hash: ed989d673d | Model: dreamshaper_7 | Denoising strength: 0.5 | Version: 5eb9f38 | Parser: Full parser | Hires upscale: 2 | Hires upscaler: Latent

Cleavage Helper

Anime models are already good at achieving the best kind of cleavage, so I expect the most value out of this version will be from helping give cleavage to smaller sized breasts.The main difference in training between this version and the regular CleavageHelper is that CleavageHelper-Anime was trained on clip skip 2 instead of 1 and with anime style images in addition to photos. This version was tested on anime models only and so the training parameters and chosen final version was determined based upon that. You can try using it on non-anime models, but it was not tested or tuned for good results on non-anime models.

Files with SHA256: ece1b0ce260658ebd5da0b688526154b577088c1d780a7e4590d1c711e789a4e