README.md

---
title: Lama-Cleaner
emoji: 🧹
colorFrom: indigo
colorTo: gray
sdk: gradio
sdk_version: 6.3.0
app_file: app.py
pinned: false
license: apache-2.0
tags:
  - image-inpainting
  - object-removal
  - lama
  - mcp-server
short_description: Remove unwanted objects from images with LaMa
---

# Lama-Cleaner: Image Inpainting

Remove unwanted objects from your images using LaMa (Large Mask Inpainting).

## Features

- **Object Removal** - Remove any unwanted object, person, or defect from images
- **LaMa Model** - Uses state-of-the-art LaMa inpainting model
- **CPU Inference** - Runs on HuggingFace Spaces free tier
- **CLI Support** - Command-line interface for batch processing

## Usage

1. Upload an image
2. Draw over the area you want to remove (white brush = mask)
3. Click "Remove Object"

## Tips

- Draw the mask slightly larger than the object for best results
- LaMa works best for small to medium sized areas
- For complex backgrounds, you may need to adjust the mask

---

## API

### Python Client

```python
from gradio_client import Client, handle_file

client = Client("Luminia/lama-cleaner")

# Note: ImageEditor data format
result = client.predict(
    editor_data={
        "background": handle_file("image.png"),
        "layers": [handle_file("mask.png")],
        "composite": None
    },
    api_name="/inpaint"
)
print(result)  # (output_image, status)
```

### REST API (curl)

```bash
# Step 1: Submit job
curl -X POST "https://luminia-lama-cleaner.hf.space/gradio_api/call/inpaint" \
  -H "Content-Type: application/json" \
  -d '{"data": [{"background": "...", "layers": [...]}]}'

# Step 2: Get result (SSE stream)
curl "https://luminia-lama-cleaner.hf.space/gradio_api/call/inpaint/{event_id}"
```

### MCP (Model Context Protocol)

This Space supports MCP for AI assistants (Claude Desktop, Cursor, VS Code).

1. Click **MCP** badge → **Add to MCP tools**
2. The `inpaint` tool becomes available

**Tool schema:**
```json
{
  "name": "inpaint",
  "parameters": {
    "editor_data": {"type": "object", "description": "ImageEditor data with background and mask layers"}
  },
  "returns": ["image", "string"]
}
```

**MCP Config:**
```json
{
  "mcpServers": {
    "lama-cleaner": {"url": "https://luminia-lama-cleaner.hf.space/gradio_api/mcp/"}
  }
}
```

---

## CLI Usage

```bash
# Inpaint with external mask
python app.py inpaint -i image.png -m mask.png -o output.png
```

**Mask format:** White (255) = area to inpaint, Black (0) = keep

---

## Credits

Based on [LaMa](https://github.com/advimman/lama) by SAIC-Moscow and [lama-cleaner](https://github.com/Sanster/lama-cleaner) by Sanster.

Paper: [Resolution-robust Large Mask Inpainting with Fourier Convolutions](https://arxiv.org/abs/2109.07161)