Yet Another Miku LoRA (YAML)

You have to unzip all of the ten '.safetensors'-files in your local 'Lora'-folder! (not just one of them)
*you can delete the .zip-file afterwards

**don't get disturbed by the name of the files, there are zero loli miku pictures used to train this model (just messed things up in the naming process...)

How to use

• SFW and NSFW possible

• Works fine with like every good anime model

• Semi-Realistic models can produce okayish results too

• Realistic probably won't work quite well

• Hairstyle, hair ornaments, eye color and face propositions are baked in

• Stable Diffusion 1.5 Model (too high resolutions end in abominations)

• Works really well with ADetailer (at least for the faces)

• Recommended settings:

  • Use LoRA strength 0.8 - 1.0

  • 27+ Steps with DPM++ 2M SDE or DPM++ 3M SDE

  • CFG scale of 6 or more (i personally prefer 8 here)

Description

Hey, this is my first LoRA and i hope you will make good use of it. I'd scrape the images from Gelbooru, used the FiftyOne Al to delete doubles and poor-quality images and the WaifuDiffusion Tagger Al (v1.4) to tag the images.

As base model i used the animefull-final-pruned-fp16.safetensors
You may find it here: https://huggingface.co/hollowstrawberry/stable-diffusion-guide/blob/main/models/animefull-final-pruned-fp16.safetensors

At this point thanks to holostrawberry for the very nice tutorial! I'd learn a lot!

See more: https://civarchive.com/models/22530/guide-make-your-own-loras-easy-and-free

Copyright

Due the fact i used public available resources without any knowledge about their licensing this LoRA is restricted to private use.