FLUX.1-dev-Finetuned-LoRA-Makoto-Shinkai

FLUX.1-dev-LoRA-Cinderella

This is a Cinderella-themed LoRA trained on FLUX.1-dev, designed to generate storybook-style fairytale illustrations inspired by the classic Cinderella tale.

<Gallery />

Trigger words

Use the phrase:
Cinderella style
to trigger the trained LoRA style in image generation.
You can extend the prompt with scene details, characters, magic effects, or emotions.

Inference

import torch
from diffusers import FluxPipeline

pipe = FluxPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", torch_dtype=torch.bfloat16)

# Updated model path
pipe.load_lora_weights("mastersubhajit/FLUX.1-dev-Finetuned-LoRA-Makoto-Shinkai", 
                       weight_name="Flux-Finetuned-LoRA-Makoto_Shinkai.safetensors")

pipe.fuse_lora(lora_scale=1.0)
pipe.to("cuda")

prompt = (
    "Cinderella style, Cinderella stepping into a glowing crystal carriage "
    "with shimmering blue butterflies swirling around her, her ball gown "
    "sparkling with magic, moonlight shining on the palace in the distance."
)

image = pipe(
    prompt,
    num_inference_steps=24,
    guidance_scale=3.5,
    width=960, height=1280,
).images[0]

image.save('cinderella.png')