> ## Documentation Index
> Fetch the complete documentation index at: https://mintlify.com/xinntao/Real-ESRGAN/llms.txt
> Use this file to discover all available pages before exploring further.

# Anime Models

> Models optimized for anime images, illustrations, and manga

Real-ESRGAN provides specialized models for anime content, trained specifically to handle the unique characteristics of anime artwork, manga, and illustrations.

## RealESRGAN\_x4plus\_anime\_6B

<Note>
  This model is optimized specifically for anime images with a smaller network architecture (6 RRDB blocks) for efficient processing while maintaining high quality.
</Note>

### Model Specifications

| Property            | Value                                                                                                                                 |
| ------------------- | ------------------------------------------------------------------------------------------------------------------------------------- |
| **Scale**           | 4x                                                                                                                                    |
| **Architecture**    | RRDBNet                                                                                                                               |
| **RRDB Blocks**     | 6 (vs 23 in general model)                                                                                                            |
| **Features**        | 64                                                                                                                                    |
| **Growth Channels** | 32                                                                                                                                    |
| **Optimization**    | Anime-specific training data                                                                                                          |
| **Download**        | [RealESRGAN\_x4plus\_anime\_6B.pth](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.2.4/RealESRGAN_x4plus_anime_6B.pth) |

### Key Features

<CardGroup cols={2}>
  <Card title="Anime-Optimized" icon="sparkles">
    Trained specifically on anime images for better line preservation and color handling
  </Card>

  <Card title="Compact Network" icon="cube">
    Only 6 RRDB blocks (vs 23 in general model) for faster processing
  </Card>

  <Card title="Sharp Lines" icon="pen">
    Preserves crisp anime line art without over-smoothing
  </Card>

  <Card title="Vibrant Colors" icon="palette">
    Maintains anime color characteristics and gradients
  </Card>
</CardGroup>

## Usage

### PyTorch Inference

```bash theme={null}
# Download the model
wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.2.4/RealESRGAN_x4plus_anime_6B.pth -P weights

# Run inference
python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i inputs
```

### NCNN (Portable Executable)

For Windows, Linux, or MacOS without Python installation:

<Steps>
  <Step title="Download NCNN Executable">
    Download the portable executable for your platform:

    * [Windows](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesrgan-ncnn-vulkan-20220424-windows.zip)
    * [Linux](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesrgan-ncnn-vulkan-20220424-ubuntu.zip)
    * [MacOS](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesrgan-ncnn-vulkan-20220424-macos.zip)
  </Step>

  <Step title="Extract and Run">
    ```bash theme={null}
    # Windows
    ./realesrgan-ncnn-vulkan.exe -i input.jpg -o output.png -n realesrgan-x4plus-anime

    # Linux/MacOS
    ./realesrgan-ncnn-vulkan -i input.jpg -o output.png -n realesrgan-x4plus-anime
    ```
  </Step>
</Steps>

## Comparisons with waifu2x

Real-ESRGAN anime model provides significant improvements over waifu2x, especially in detail preservation and artifact reduction.

<Note>
  Comparison with waifu2x using `-n 2 -s 4` settings
</Note>

### Visual Comparison Examples

<img src="https://raw.githubusercontent.com/xinntao/public-figures/master/Real-ESRGAN/cmp_realesrgan_anime_1.png" alt="Anime comparison 1" />

<img src="https://raw.githubusercontent.com/xinntao/public-figures/master/Real-ESRGAN/cmp_realesrgan_anime_2.png" alt="Anime comparison 2" />

<img src="https://raw.githubusercontent.com/xinntao/public-figures/master/Real-ESRGAN/cmp_realesrgan_anime_3.png" alt="Anime comparison 3" />

<img src="https://raw.githubusercontent.com/xinntao/public-figures/master/Real-ESRGAN/cmp_realesrgan_anime_4.png" alt="Anime comparison 4" />

<img src="https://raw.githubusercontent.com/xinntao/public-figures/master/Real-ESRGAN/cmp_realesrgan_anime_5.png" alt="Anime comparison 5" />

### Advantages Over waifu2x

<CardGroup cols={2}>
  <Card title="Better Detail Recovery" icon="magnifying-glass-plus">
    Recovers fine details in hair, clothing textures, and background elements
  </Card>

  <Card title="Reduced Artifacts" icon="shield-check">
    Fewer compression artifacts and smoother gradients
  </Card>

  <Card title="Line Preservation" icon="pen">
    Maintains crisp line art without introducing halos or blur
  </Card>

  <Card title="Color Accuracy" icon="palette">
    Better preservation of original anime color characteristics
  </Card>
</CardGroup>

## Best Practices

### For Anime Images

<AccordionGroup>
  <Accordion title="Line Art and Manga">
    The anime model excels at preserving clean line art:

    ```bash theme={null}
    python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i manga_pages
    ```
  </Accordion>

  <Accordion title="Colored Illustrations">
    Works great for colored anime artwork:

    ```bash theme={null}
    python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i anime_art
    ```
  </Accordion>

  <Accordion title="Screenshots">
    Ideal for upscaling anime screenshots:

    ```bash theme={null}
    python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i screenshots
    ```
  </Accordion>

  <Accordion title="Large Images">
    Use tiling for large manga pages or illustrations:

    ```bash theme={null}
    python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i inputs -t 512
    ```
  </Accordion>
</AccordionGroup>

### Command Options

```bash theme={null}
# Basic usage
python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i inputs -o outputs

# With tiling (for large images)
python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i inputs -t 512

# Custom output scale (2x instead of 4x)
python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i inputs -s 2

# Save as PNG
python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i inputs --ext png
```

## Model Architecture Comparison

### Anime vs General Model

| Feature           | RealESRGAN\_x4plus\_anime\_6B | RealESRGAN\_x4plus |
| ----------------- | ----------------------------- | ------------------ |
| **RRDB Blocks**   | 6                             | 23                 |
| **Parameters**    | Fewer                         | More               |
| **Speed**         | Faster                        | Slower             |
| **Memory Usage**  | Lower                         | Higher             |
| **Training Data** | Anime-specific                | General images     |
| **Best For**      | Anime, manga, illustrations   | Natural photos     |

<Note>
  The anime model's smaller architecture (6 blocks) makes it approximately **3-4x faster** than the general model while maintaining high quality for anime content.
</Note>

## Discriminator Model

<Accordion title="For Fine-tuning">
  If you need to fine-tune the anime model on your own dataset:

  | Property                    | Value                                                                                                                                            |
  | --------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------ |
  | **Discriminator**           | [RealESRGAN\_x4plus\_anime\_6B\_netD.pth](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.2.4/RealESRGAN_x4plus_anime_6B_netD.pth) |
  | **Corresponding Generator** | RealESRGAN\_x4plus\_anime\_6B                                                                                                                    |

  This discriminator model is used during GAN training and is not needed for inference.
</Accordion>

## Performance Tips

<CardGroup cols={2}>
  <Card title="Batch Processing" icon="layer-group">
    Process multiple images efficiently:

    ```bash theme={null}
    python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i input_folder
    ```
  </Card>

  <Card title="GPU Memory" icon="memory">
    Use tiling for large images:

    ```bash theme={null}
    python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i inputs -t 400
    ```
  </Card>

  <Card title="Custom Scale" icon="expand">
    Output at different scales:

    ```bash theme={null}
    python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i inputs -s 2
    ```
  </Card>

  <Card title="Format Control" icon="image">
    Control output format:

    ```bash theme={null}
    python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i inputs --ext png
    ```
  </Card>
</CardGroup>

## Next Steps

<CardGroup cols={2}>
  <Card title="Video Models" icon="film" href="/models/video-models">
    Learn about anime video upscaling
  </Card>

  <Card title="General Models" icon="image" href="/models/general-models">
    Explore models for natural images
  </Card>
</CardGroup>