---
license: creativeml-openrail-m
tags:
  - flux
  - lora
  - kontext
  - avatar
  - image-to-image
  - style-transfer
  - na'vi
  - stable-diffusion
  - art
  - character-design
  - ai-generated
inference: true
---

# FLUX.1-Kontext-Dev Na'vi Style LoRA 🎨🧞‍♂️

This is a LoRA adapter trained on top of the [`black-forest-labs/FLUX.1-Kontext-dev`](https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev) model to transform images into *Avatar-style Na'vi characters*, inspired by the visuals from James Cameron's *Avatar* films.

## 📸 Example Outputs

| Input Image | Stylized Output |
|-------------|-----------------|
| ![input](./examples/input.jpg) | ![output](./examples/output.jpg) |

> *(Add your own examples by placing them in an `examples/` folder in the repo)*

---

## 🧠 Model Details

- **Base Model**: FLUX.1-Kontext-dev (custom diffusion)
- **LoRA Rank**: 4 / 8 (depending on your version)
- **Training Resolution**: 512x512
- **Target Style**: Na'vi characters (blue skin, elongated facial features, glowing eyes, etc.)
- **Architecture**: LoRA fine-tuning with Stable Diffusion compatible UNet blocks.

---

## 💻 Usage

### In Diffusers

```python
from diffusers import StableDiffusionPipeline
from peft import PeftModel
import torch

# Load the base model
pipe = StableDiffusionPipeline.from_pretrained("black-forest-labs/FLUX.1-Kontext-dev", torch_dtype=torch.float16).to("cuda")

# Load the LoRA adapter
pipe.unet.load_attn_procs("your-username/flux-kontext-na-vi-lora")

# Use the model
prompt = "A close-up portrait of a Na'vi warrior in a glowing forest, cinematic lighting"
image = pipe(prompt).images[0]
image.save("output.png")