---
base_model: stabilityai/stable-diffusion-3.5-large
library_name: diffusers
tags:
- text-to-image
- diffusers-training
- diffusers
- lora
- sd3
- sd3-diffusers
- template:sd-lora
instance_prompt: >-
  Ghibli style [character description] with [distinctive features], [action or
  pose], [environment or background], [lighting or atmosphere], [additional
  details]
license: other
license_name: stabilityai-ai-community
license_link: https://huggingface.co/stabilityai/stable-diffusion-3.5-large/blob/main/LICENSE.md
widget:
- text: >-
    Ghibli style futuristic stormtrooper with glossy white armor and a sleek
    helmet, standing heroically on a lush alien planet, vibrant flowers blooming
    around, soft sunlight illuminating the scene, a gentle breeze rustling the
    leaves
  output:
    url: images/stormtrooper.jpg
datasets:
- alvarobartt/ghibli-characters
pipeline_tag: text-to-image
---

# Studio Ghibli Characters - SD 3.5 Large LoRA

<Gallery />

## Model description

This repository contains the LoRA adapter for [StableDiffusion 3.5 Large](https://huggingface.co/stabilityai/stable-diffusion-3.5-large), fine-tuned using [Diffusers + Dreambooth](https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth_lora_sd3.py)
with Studio Ghibli images free-to-use downloaded manually from https://ghibli.jp and hosted at [alvarobartt/ghibli-characters](https://huggingface.co/datasets/alvarobartt/ghibli-characters).

## Prompt Template

You should use the following template (defined when annotating the images with the captions) to trigger the image generation:

"Ghibli style [character description] with [distinctive features], [action or pose], [environment or background], [lighting or atmosphere], [additional details]"

## Inference with `diffusers`

> [!NOTE]
> Requires `diffusers` 0.31.0 or higher, see the [release notes](https://github.com/huggingface/diffusers/releases/tag/v0.31.0).

```python
import torch
from diffusers import DiffusionPipeline

model_id = "stabilityai/stable-diffusion-3.5-large"
adapter_id = "alvarobartt/ghibli-characters-sd3.5-lora"

pipeline = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.bfloat16)
pipeline.load_lora_weights(adapter_id)
pipeline.to("cuda")

prompt = (
    "Ghibli style futuristic stormtrooper with glossy white armor and a sleek helmet,"
    " standing heroically on a lush alien planet, vibrant flowers blooming around, soft"
    " sunlight illuminating the scene, a gentle breeze rustling the leaves"
)

image = pipeline(
    prompt=prompt,
    num_inference_steps=30,
    width=1024,
    height=768,
    guidance_scale=3.5,
).images[0]

image.save("ghibli.png", format="PNG")
```

## Disclaimer

This fine-tune is for **personal-use only**, with **no-commercial purposes** as stated within the licensing.

The StableDiffusion 3.5 Large fine-tunes fall under the same license as StableDiffusion 3.5 Large i.e. https://huggingface.co/stabilityai/stable-diffusion-3.5-large/blob/main/LICENSE.md;
and the Studio Ghibli dataset is released with a custom non-commercial license based on the interpretation of the findings on their website.