Please use supermerger project: https://github.com/hako-mikan/sd-webui-supermerger
It has improved version of cosine similarity with awesome UI
TLDR; Mixing models by cosine similarity
python3 weightedsim.py openjourney-v2.ckpt EimisAnimeDiffusion_1-0v.ckpt
25 year old woman, white top, blue shorts, adorable face, piercing eyes, resting mouth, intricate necklace, short hair, looking away, standing next to beach, sunset, palm tree
Negative prompt: deformed, bad anatomy, disfigured, mutation, extra limb, ugly, poorly drawn hands, missing limb, floating limbs, disconnected limbs, malformed hands, blurry, ((((mutated hands and fingers)))), distorted hands, amputation, missing hands, double hands, watermark, censored, black and white, sepia, zombie
Steps: 28, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 0, Size: 576x832, Model hash: 628090e8bb, Model: merged-0.5, ENSD: 31337
Please stop asking me how to run this and how to work with it. I wrote it without knowing Python. This is my second and hopefully last Python script.
DEw2DR8C7BnF8GgcrfTzUjSnGkuMeJhg83