「fuduki_mix v2.0」が公開されています。過去のモデルと比較してみました。

結果

左から v1.0 → v1.5 → v2.0 です。







今回使用したプロンプトでは、キリっとしたかっこいい女性から綺麗な女性に変わってきている気がします。

Pythonスクリプト

プロンプトはこちらからそのまま使わせてもらいました。

from diffusers import DiffusionPipeline, DPMSolverMultistepScheduler
import torch
import os
import cv2
import numpy as np
import time

n_samples = 5
model_safetensors = {
    "v1.0": "model/fudukiMix_v10",
    "v1.5": "model/fudukiMix_v15",
    "v2.0": "model/fudukiMix_v20"
}

def txt2img(model_name:str) -> None:
    model_path = model_safetensors[model_name]
    pipe = DiffusionPipeline.from_pretrained(
        model_path,
        custom_pipeline = "lpw_stable_diffusion_xl",
        torch_dtype=torch.float16,
        variant="fp16"
    )

    pipe.scheduler = DPMSolverMultistepScheduler.from_config(
        pipe.scheduler.config,
        algorithm_type="sde-dpmsolver++",
        use_karras_sigmas=True
    )
    
    pipe.to("cuda")

    prompt = "japanese woman, (close-up:2), (natural lighting), wavy hair, forehead, (dark brown eyes), (from side), (downward slanting eyes), pouty, (white sweater), (dyanmic posing), (see-through curtain), (bright room)"
    n_prompt = "(cleavage:2), (illustration), 3d, 2d, painting, cartoons, sketch, watercolor, monotone, (kimono), (crossed eyes), (strabismus)"
    
    os.makedirs(model_name, exist_ok=True)

    for i in range(n_samples):
        seed = 100000 * (i + 1)
        generator = torch.manual_seed(seed)
        image = pipe(
            prompt=prompt,
            negative_prompt=n_prompt,
            generator=generator,
            guidance_scale=5.0,
            num_inference_steps = 35,
            width=896,
            height=1152
        ).images[0]

        image.save(os.path.join(model_name, f"{i}.png"))

def stack() -> None:
    os.makedirs("stack_image", exist_ok=True)
    print(" -> ".join(model_safetensors.keys()))
    for i in range(n_samples):
        images_list = []
        for key in model_safetensors.keys():
            images_list.append(cv2.imread(os.path.join(key, f"{i}.png")))
        stack_image = np.hstack(images_list)
        cv2.imwrite(os.path.join("stack_image", f"{i}.png"), stack_image)

if __name__ == "__main__":
    start = time.time()
    
    for model in model_safetensors:
        txt2img(model)
    stack()
    
    end = time.time()
    print(f"処理時間: {end - start:.5f}秒")