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

# AI image enhancement

> Use AI-powered upscaling, descreening, and artifact removal to improve comic image quality

OpenComic includes AI-powered image enhancement features that can dramatically improve the quality of low-resolution or compressed comic images. Using the `opencomic-ai-bin` package, you can upscale images, remove compression artifacts, and eliminate screening patterns.

## AI features overview

<CardGroup cols={3}>
  <Card title="Upscaling" icon="maximize">
    Increase image resolution 2-4× while preserving details
  </Card>

  <Card title="Descreen" icon="filter">
    Remove halftone screening patterns from scanned comics
  </Card>

  <Card title="Artifact removal" icon="broom">
    Clean up compression artifacts and noise
  </Card>
</CardGroup>

## AI upscaling

Upscaling uses machine learning models to intelligently increase image resolution beyond simple interpolation.

### Features

* **Scale factors**: 2×, 3×, or 4× original resolution
* **Auto-scale**: Automatically determines optimal scale based on target resolution
* **Noise reduction**: Optional noise level adjustment (0-3)
* **Multiple models**: Choose from various AI models optimized for speed or quality
* **Megapixel limit**: Set maximum input size to control performance

### Available models

<Tabs>
  <Tab title="RealCUGAN">
    Default upscaling model

    **Scales**: 2×, 3×, 4×

    **Noise levels**: -1, 0, 1, 2, 3

    **Speed**: Fast to Medium

    **Best for**: General purpose upscaling with good quality-speed balance
  </Tab>

  <Tab title="Lite models">
    OpenComic AI Upscale Lite

    **Scales**: 2×, 4×

    **Speed**: Very Fast

    **Best for**: Quick upscaling when performance is priority
  </Tab>
</Tabs>

### Configuration

<Steps>
  <Step title="Enable upscaling">
    Navigate to Reading Settings → AI Enhancement → Upscale

    Toggle "Active" to enable AI upscaling.
  </Step>

  <Step title="Select model">
    Choose an AI model based on your performance needs:

    * **RealCUGAN**: Best quality (default)
    * **Lite models**: Faster processing

    Each model displays its relative speed (Very Fast, Fast, Medium, Slow, Very Slow).
  </Step>

  <Step title="Set scale factor">
    Choose the scale multiplier:

    * **2×**: Double resolution (moderate quality gain)
    * **3×**: Triple resolution (significant improvement)
    * **4×**: Quadruple resolution (maximum quality)

    Higher scales take longer to process.
  </Step>

  <Step title="Configure auto-scale" icon="wand-magic">
    Enable auto-scale to let OpenComic automatically choose the best scale factor:

    * Calculates based on target megapixels
    * Prevents excessive upscaling
    * Optimizes for your display resolution
  </Step>

  <Step title="Adjust noise reduction" icon="sliders">
    For models that support it, set noise reduction level (0-3):

    * **-1**: No noise reduction
    * **0**: Minimal noise reduction
    * **1-3**: Increasing noise reduction strength
  </Step>

  <Step title="Set megapixel limit">
    Configure maximum input image size:

    ```javascript theme={null}
    maxMegapixels: 3.0  // Only upscale images up to 3MP
    ```

    Images larger than this limit won't be upscaled.
  </Step>
</Steps>

### Auto-scale behavior

When auto-scale is enabled:

```javascript theme={null}
// Calculation formula
scale = min(max(round((√maxMegapixels / √currentMegapixels) × 2), 2), 4)

// Example: 1MP image with 3MP target
// scale = min(max(round((√3 / √1) × 2), 2), 4)
// scale = min(max(round(3.46), 2), 4) = 3
```

### Performance considerations

<Warning>
  AI upscaling is computationally intensive. Processing time depends on:

  * Image resolution
  * Selected model
  * Scale factor
  * Your CPU/GPU capabilities
</Warning>

**Optimization tips:**

* Use lower scale factors for large images
* Enable auto-scale to prevent over-processing
* Set appropriate megapixel limits
* Choose faster models for real-time reading
* Pre-process comics in advance for better experience

## Descreening

Descreening removes halftone patterns (dots/screens) commonly found in scanned print comics.

### What is screening?

Traditional print comics use halftone screening:

* **Halftone dots**: Create gradients and colors through dot patterns
* **Moiré patterns**: Interference patterns from scanning
* **Screen angles**: Misaligned screen grids causing artifacts

### Models

<Tabs>
  <Tab title="Hard Descreen">
    OpenComic AI Descreen Hard Lite

    **Strength**: High

    **Speed**: Medium

    **Best for**: Heavy screening patterns, older scans
  </Tab>

  <Tab title="Soft Descreen">
    OpenComic AI Descreen Soft Lite

    **Strength**: Moderate

    **Speed**: Medium

    **Best for**: Light screening, preserving fine details
  </Tab>
</Tabs>

### Configuration

<Steps>
  <Step title="Enable descreening">
    Navigate to Reading Settings → AI Enhancement → Descreen

    Toggle "Active" to enable descreening.
  </Step>

  <Step title="Select model">
    Choose between hard or soft descreening:

    * **Hard**: Aggressive screening removal
    * **Soft**: Gentle screening reduction
  </Step>

  <Step title="Apply">
    Changes apply immediately to your reading view.
  </Step>
</Steps>

### When to use descreening

<CardGroup cols={2}>
  <Card title="Good candidates" icon="check">
    * Scanned print comics
    * Visible halftone patterns
    * Moiré interference
    * Older, low-quality scans
  </Card>

  <Card title="Skip descreening" icon="xmark">
    * Digital-first comics
    * Clean, high-resolution scans
    * Vector/redrawn artwork
    * Already descreened images
  </Card>
</CardGroup>

## Artifact removal

Remove compression artifacts, blocking, and noise from JPEG and other lossy formats.

### Common artifacts

* **JPEG blocks**: 8×8 pixel compression blocks
* **Mosquito noise**: Ringing around edges
* **Color banding**: Posterization in gradients
* **General noise**: Random pixel variations

### Model

OpenComic AI Artifact Removal Lite

**Speed**: Medium

**Strength**: Balanced artifact removal

**Best for**: Compressed comic files (CBR, CBZ with JPEG images)

### Configuration

<Steps>
  <Step title="Enable artifact removal">
    Navigate to Reading Settings → AI Enhancement → Artifact Removal

    Toggle "Active" to enable.
  </Step>

  <Step title="Select model">
    Currently one model available (more coming in future updates).
  </Step>

  <Step title="Apply">
    Artifact removal processes images automatically while reading.
  </Step>
</Steps>

### Effectiveness

Artifact removal works best on:

* **Moderate compression**: Visible but not extreme artifacts
* **JPEG images**: Common in comic archives
* **Color images**: Removes color fringing and banding
* **Edge artifacts**: Cleans up ringing and halos

<Note>
  Heavily damaged or extremely compressed images may not fully recover. Artifact removal works best as preventive enhancement.
</Note>

## Combining enhancements

You can enable multiple AI enhancements simultaneously:

### Processing order

```javascript theme={null}
1. Artifact Removal  // Clean up compression first
2. Descreen         // Remove halftone patterns
3. Upscale          // Scale up clean image
```

This order ensures:

* Artifacts don't get amplified by upscaling
* Screening patterns removed before scaling
* Final upscaling produces cleanest result

### Performance impact

<Warning>
  Enabling all three features significantly increases processing time. Consider:

  * Enable only needed enhancements
  * Use faster models
  * Process important comics in advance
  * Disable for non-problematic content
</Warning>

## Model selection guide

<AccordionGroup>
  <Accordion title="For speed (real-time reading)">
    * **Upscale**: Lite models, scale 2×
    * **Descreen**: Soft model
    * **Artifact removal**: Lite model
    * **Megapixel limit**: 1.5-2.0 MP
  </Accordion>

  <Accordion title="For quality (pre-processing)">
    * **Upscale**: RealCUGAN, scale 3-4×
    * **Descreen**: Hard model
    * **Artifact removal**: Enabled
    * **Megapixel limit**: 3.0+ MP
  </Accordion>

  <Accordion title="For old scans">
    * **Upscale**: RealCUGAN with noise reduction
    * **Descreen**: Hard model (essential)
    * **Artifact removal**: Enabled
    * **Noise level**: 2-3
  </Accordion>

  <Accordion title="For digital comics">
    * **Upscale**: Only if low-res
    * **Descreen**: Disabled
    * **Artifact removal**: Only if compressed
  </Accordion>
</AccordionGroup>

## Image interpolation

In addition to AI upscaling, OpenComic supports traditional interpolation methods:

### Available methods

<Tabs>
  <Tab title="Upscaling">
    * **lanczos3**: Best quality (default)
    * **lanczos2**: Good quality, faster
    * **mitchell**: Balanced
    * **cubic**: Smooth gradients
    * **linear**: Fast, acceptable quality
    * **nearest**: Fastest, pixelated
  </Tab>

  <Tab title="Downscaling">
    * **lanczos3**: Best quality (default)
    * **mitchell**: Good quality
    * **cubic**: Smooth results
    * **linear**: Fast
    * **nearest**: Fastest
  </Tab>
</Tabs>

### Configuration

Set interpolation method in settings:

* **Upscaling method**: Used when enlarging images
* **Downscaling method**: Used when shrinking images

<Note>
  AI upscaling bypasses interpolation methods. Set these for non-AI image scaling.
</Note>

## Caching and storage

### AI processing cache

Processed images are cached:

```javascript theme={null}
// Cache location
tempFolder/ai-upscale/{sha}/

// Cached files
- Original image hash
- Processed image
- Metadata
```

### Cache benefits

* **Skip reprocessing**: Already-processed images load instantly
* **Per-comic caching**: Each comic has its own cache folder
* **Persistent**: Cache survives app restarts

### Cache management

<AccordionGroup>
  <Accordion title="Automatic cleanup">
    OpenComic automatically manages cache size, removing old entries as needed.
  </Accordion>

  <Accordion title="Manual clearing">
    Clear AI cache from settings to free disk space or force reprocessing.
  </Accordion>
</AccordionGroup>

## Per-page configuration

OpenComic allows different AI settings per page type:

### Page-specific settings

```javascript theme={null}
readingPagesConfig: {
  single: {
    readingAi: {
      upscale: { active: true, scale: 3 },
      descreen: { active: true },
      artifactRemoval: { active: true }
    }
  },
  double: {
    readingAi: {
      upscale: { active: false },  // Disable for double pages
      descreen: { active: true },
      artifactRemoval: { active: true }
    }
  },
  webtoon: {
    readingAi: {
      upscale: { active: true, scale: 2 },  // Lower scale for webtoons
      descreen: { active: false },
      artifactRemoval: { active: true }
    }
  }
}
```

### Benefits

* **Optimize per format**: Different settings for different page types
* **Performance tuning**: Disable intensive features for large pages
* **Quality control**: Maximum enhancement where it matters most

## Troubleshooting

### Slow processing

<Steps>
  <Step title="Reduce scale factor">
    Lower upscale multiplier (4× → 2×)
  </Step>

  <Step title="Use faster models">
    Switch to Lite models for better performance
  </Step>

  <Step title="Lower megapixel limit">
    Process only smaller images
  </Step>

  <Step title="Disable features">
    Turn off unneeded enhancements
  </Step>
</Steps>

### Quality issues

**Upscaling artifacts:**

* Try different noise reduction levels
* Switch models
* Lower scale factor

**Over-processing:**

* Disable unnecessary features
* Use softer descreen model
* Reduce noise reduction

### Crashes or errors

1. **Check AI models**: Ensure models are properly installed
2. **Verify system resources**: Ensure sufficient RAM
3. **Update OpenComic**: Latest version has performance improvements
4. **Clear cache**: Remove corrupted cached files

***

<Card title="Next: Customization" icon="palette" href="/advanced/customization">
  Customize themes, shortcuts, filters, and tap zones
</Card>