A model Made to give The best Anime realism outputs with a few parts of regular models.

Link to The vae used: -> Here <-

Vanilla is the newest version of MasterAnime, And will most likely be the last for a while becuase I have not been finding inexigent improvements in the models, But all the Models I was working on are in my HuggingFace repo And if you like it enough I might post that version here.

Golden Cherry, is The successor of Vanilla, will have had somewhere around 1.5Kish models in total merged into it, and is currently being tested, if it's good I will release it.

V1 is a merge of 128~ Models (Using weight sum and normal calc mode) V2 is a Merge of 64 new models + V1 and a couple test versions. V3 is basically a merge of V2 plus some newer models, I think about 64~ but it could be more. V4 is a merge of V3 plus some newer models. V4.5 is a merge meant to replicate the base style of Anything V4.5 but with all around better prompting and anatomy, although now that I have tested it, It is better than V4 BUT V4.5 has problems with eyes mostly.

HERE IS A BASIC GUIDE:

Use Adetailer, Use what ever settings you want; I suggest DPM++ 2M SDE Heun Karras and 38 steps. for upscaler, just any anime style upscaler will do.

For prompting here is a basic prompt which works:

(masterpiece, highest quality, 8k, 16k, high quality, best quality:1.3, absurdres, highres, digital art, intricate), <lora:GoodHands-vanilla:1>

and the negative:

(worst quality:2),(low quality:2),(normal quality:2),lowres,((monochrome)),((grayscale)),watermark, blurred, bad face, child, shota, lolli, old, ugly, bad anatomy,(long body :1.3),(mutation, poorly drawn :1.2),(extra fingers:1.2), ( extra fingers, deformed hands), (poor quality, bad quality:1.35), <lora:neg4all_bdsqlsz_V3.5:1>, FastNegativeV2, EasyNegativeV2, NGH,

Note you will need the embeds + Locoris if you want it too look really good... also you can add nsfw to the negative and it will try to not make nsfw stuff.

描述:

Copyright LOLOLOLOLOLOL

训练词语: