A Lunar Dream Beta

This is a new trained model that can produce good quality anime girls normally but has a secret trick up its sleeve. It has many styles trained into the model that you can recall. As of V1 it has 40 art styles. I will be making V2 when I get 100 artists to include. This is going to take a hile to find 100 artists, so if you can donate a dataset, I'd be appreciative. ???? images at the back.

• Can it do men? Yes, it actually has a lot of men in the dataset

• Can it do ????? Yes, You will have to prompt what you want to not get jank though. This model is NOT overfit like the models you are used to. If you don't prompt the pose, you will probably get jank. Some styles won't proc and instead it'll mostly default to the main ???? artist.

• Do I need to use an artist name? No, it has its own unique style if you don't prompt in an artist. No other model should look like this. It's a decently clean anime style.

• Can you mix artists? YES. Results will be mixed depending on who you mix obviously, but you can get really nice results.

• Does it produce jank? Yes. You need to be careful with your prompts as this is not as overfit as you are used to. It will listen to your prompts too good sometimes, so make sure you don't double up on the same thing etc.

You can add the lunachan tag to your prompt and it will slightly alter the generations. It typically improves it a tad. Some styles will work better if you combine the name and style"artistname, artistname style". I will improving this in V2 by having every artist work this way.

I have created a handy umi Ai wildcard to generate a random style. I will include it on my huggingface when I get a chance. But for now, you can toss this in your prompt if you want to random it. Or make your own umi ai .txt with it.

{82Pigeon style|akai Sashimi style|Algae Makku style|belko style|braveking style|chocofing style|Dako style|esuna style|evaaviva style|fengweic style|gomennasai style|hitokomoru style|ibukichi style|ishikei style|jidao style|junc style|Khyle style|kyuyongeom style|lam style|lovecacao style|mamamba style|matcha style|mdastarou style|melanbread style|melodyoflost style|MikaPikazo style|morry style|murata Range style|naruko Hanaharu style|navigavi style|rei 17 style|rupid style|scottie style|spacezin style|tahra style|timrougami style|tira style|Tsutomu style|Uekura Eku style|xiayu style|yi1zhi style|yomu style|Yoneyama style|yoshio style|yueko style}

As always, no lora etc were used to produce the same images. This is pure the model doing the artist. I personally like using the Blessed.Vae as it produces vibrant images but not as bright as the WD vae. I think it's on huggingface somewhere.

If you use this as a merge, can you do me the kindness of mentioning me/linking here?

If you have an artist you want me to add, message me! If you have a dataset you can send, me, send me it!