---
title: flux-2-klein-4B-outpaint-lora
canonical_url: "https://www.modelscope.cn/models/RealmSky/flux-2-klein-4B-outpaint-lora"
md_url: "https://www.modelscope.cn/models/RealmSky/flux-2-klein-4B-outpaint-lora.md"
repository: RealmSky/flux-2-klein-4B-outpaint-lora
last_updated: 2026-02-01
license: apache-2.0
pipeline_tag: image-to-image
tasks:
  - image-to-image
base_model:
  - black-forest-labs/FLUX.2-klein-4B
base_model_relation: adapter
parameters: 38.0M
tensor_type:
  - F32
library_name:
  - lora
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
supports_inference: img2img
downloads: 209
stars: 3
tags:
  - flux
  - flux-2-klein
  - flux-lora
  - lora
  - outpainting
  - image-extension
  - green-screen
  - inpainting
  - image-to-image
  - fal-ai
  - "template:diffusion-lora"
---

# flux-2-klein-4B-outpaint-lora

> flux-2-klein-4B-outpaint-lora - RealmSky 在 ModelScope 开源的模型。🖼️ Flux Green Screen Outpainting LoRA

RealmSky/flux-2-klein-4B-outpaint-lora 是 ModelScope 魔搭社区上的 38.0M 参数image-to-image模型，采用 apache-2.0 许可，基于 black-forest-labs/FLUX.2-klein-4B 构建，并支持在线推理（img2img）。

- **Repository**: RealmSky/flux-2-klein-4B-outpaint-lora
- **License**: apache-2.0
- **Tasks**: image-to-image
- **Parameters**: 38.0M
- **Base model**: black-forest-labs/FLUX.2-klein-4B
- **Online inference**: img2img
- **Tags**: flux, flux-2-klein, flux-lora, lora, outpainting, image-extension, green-screen, inpainting, image-to-image, fal-ai, template:diffusion-lora
- **Downloads**: 209
- **Stars**: 3
- **Last updated**: 2026-02-01

Source: https://www.modelscope.cn/models/RealmSky/flux-2-klein-4B-outpaint-lora

---

<Gallery />

<div align="center">

# 🖼️ Flux Green Screen Outpainting LoRA

### Extend images by filling green border areas with AI-generated content

[![Try on fal.ai](https://img.shields.io/badge/🚀_Try_it_Live-fal.ai-FF6B35?style=for-the-badge)](https://fal.ai/models/fal-ai/flux-2/klein/4b/base/edit/lora)
[![License](https://img.shields.io/badge/License-Apache_2.0-green?style=for-the-badge)](https://opensource.org/licenses/Apache-2.0)

<img src="images/banner.jpg" alt="Outpaint LoRA Banner" width="100%">

</div>

---

## 🎯 What does this model do?

This LoRA fills **green border areas** around an image with contextually appropriate AI-generated content. Simply add a green border to any image, and the model will extend the scene naturally.
```
┌─────────────────────────────────────┐
│ ░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░ │  Green areas
│ ░░░┌───────────────────────┐░░░░░░░ │  get filled
│ ░░░│                       │░░░░░░░ │  with matching
│ ░░░│    Original Image     │░░░░░░░ │  content
│ ░░░│                       │░░░░░░░ │
│ ░░░└───────────────────────┘░░░░░░░ │
│ ░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░ │
└─────────────────────────────────────┘
```

Perfect for **extending photos**, **creating wider compositions**, and **generating seamless backgrounds**.

---

## 🖼️ Examples

<table>
<tr>
<th>Input (with green border)</th>
<th>Output (extended image)</th>
</tr>
<tr>
<td><img src="examples/example1_input.jpg" width="300"></td>
<td><img src="examples/example1_output.jpg" width="300"></td>
</tr>
<tr>
<td><img src="examples/example2_input.jpg" width="300"></td>
<td><img src="examples/example2_output.jpg" width="300"></td>
</tr>
<tr>
<td><img src="examples/example3_input.jpg" width="300"></td>
<td><img src="examples/example3_output.jpg" width="300"></td>
</tr>
<tr>
<td><img src="examples/example4_input.jpg" width="300"></td>
<td><img src="examples/example4_output.jpg" width="300"></td>
</tr>
</table>

---

## 🚀 Quick Start

### Prompt
```
Fill the green spaces according to the image
```

### Green Border Requirements
- Use pure green: RGB(0, 255, 0) or #00FF00
- Border can be any thickness (5%-25% of image dimension works well)
- Each side can have different border widths

---

## 💻 Usage

### Try it Live on fal.ai

▶️ **[Open Playground](https://fal.ai/models/fal-ai/flux-2-klein/4b/base/edit/lora?share=528658ce-ed9c-46be-8e2b-815fbc4d14a6)**

### With fal.ai SDK
```python
import fal_client

def on_queue_update(update):
    if isinstance(update, fal_client.InProgress):
        for log in update.logs:
           print(log["message"])

result = fal_client.subscribe(
    "fal-ai/flux-2/klein/4b/base/edit/lora",
    arguments={
        "prompt": "Fill the green spaces according to the image",
        "model_name": None,
        "loras": [{
            "path": "https://huggingface.co/fal/flux-2-klein-4B-outpaint-lora/resolve/main/flux-outpaint-lora.safetensors",
            "scale": 1.1
        }],
        "embeddings": [],
        "image_urls": ["https://your-image-with-green-border.png"]
    },
    with_logs=True,
    on_queue_update=on_queue_update,
)
print(result)
```

---

## 📦 Model Files

| File | Use Case |
|------|----------|
| `flux-outpaint-lora.safetensors` | fal.ai |
| `LyNiaZ53Tudg0J6sT8Xbx_pytorch_lora_weights_comfy_converted.safetensors` | ComfyUI |

---

## 🎓 Training Details

<details>
<summary>Click to expand</summary>

### Dataset
- **Size**: 1000 image pairs
- **Content**: Diverse subjects including:
  - **Landscapes**: mountains, beaches, forests, deserts, cities
  - **People**: portraits, groups, activities
  - **Animals**: pets, wildlife, birds
  - **Food**: dishes, ingredients, table settings
  - **Architecture**: buildings, interiors, landmarks
  - **Art styles**: paintings, illustrations, abstract
  - **Objects**: vehicles, electronics, furniture
  - **Nature**: weather, space, underwater
- **Border variation**: Random green borders (5%-25% of each side)
- **Aspect ratios**: Various (1:1, 16:9, 9:16, 4:3, 3:4, etc.)

### Training
- **Base Model**: FLUX.2-Klein 4B
- **Platform**: [fal.ai](https://fal.ai)
- **Method**: LoRA training

</details>

---

## 🎮 Use Cases

- **Photo Extension**: Expand cropped photos to wider compositions
- **Background Generation**: Create seamless backgrounds for subjects
- **Aspect Ratio Conversion**: Convert portrait to landscape or vice versa
- **Creative Composition**: Add environmental context to isolated subjects

---

## ⚠️ Limitations

- Works best with pure green (#00FF00) borders
- Complex scenes may require multiple passes
- Very large border areas may produce less coherent results

---

## 📄 License

[Apache 2.0](https://opensource.org/licenses/Apache-2.0)

---

<div align="center">

### Created by [@ilkerzgi](https://huggingface.co/ilkerzgi) • Trained with [fal.ai](https://fal.ai)


[![Try on fal.ai](https://img.shields.io/badge/🚀_Try_it_Live-fal.ai-FF6B35?style=for-the-badge)](https://fal.ai/models/fal-ai/flux-2/klein/4b/base/edit/lora/)

</div>