Smart Chunking

Smart Chunking

Smart Chunking is ImageChunker’s signature feature — an intelligent algorithm that analyzes your images to find optimal split points.

How It Works

Unlike simple grid-based splitting, Smart Chunking:

1. Scans image content — Analyzes pixel patterns and content density
2. Detects boundaries — Identifies natural break points (blank spaces, scene changes)
3. Avoids awkward cuts — Won’t split through text, faces, or key elements
4. Optimizes distribution — Balances chunk sizes while respecting content

The Algorithm

Smart Chunking uses a multi-pass approach:

Pass 1: Edge Detection
├── Scan horizontal slices for content changes
├── Identify high-contrast boundaries
└── Mark potential split candidates

Pass 2: Content Analysis
├── Detect text regions (avoid mid-word splits)
├── Identify faces and figures
└── Score each candidate by "splitability"

Pass 3: Optimization
├── Balance chunk sizes (prefer even distribution)
├── Respect minimum/maximum chunk heights
└── Select optimal split points

Settings

Sensitivity

Controls how aggressively Smart Chunking looks for boundaries:

Level	Behavior	Best For
Low	Fewer, larger chunks	Dense content
Medium	Balanced splitting	Most images
High	More, smaller chunks	Varied content

Minimum Chunk Height

Prevents chunks from becoming too small:

• Default: 200px
• Range: 50-1000px
• Tip: Set higher for images with less content variation

Content Detection

Enable or disable specific detection:

• Text Detection: Avoid splitting text mid-line
• Face Detection: Keep faces intact
• Object Detection: Preserve key subjects

When to Use Smart Chunking

Ideal Cases

✅ Screenshots with mixed content

• Text sections followed by images
• UI elements of varying sizes

✅ Comic and manga pages

• Panel boundaries provide natural splits
• Dialogue should stay together

✅ Infographics

• Sections often have natural breaks
• Headers indicate split points

✅ Photos with subjects

• Portraits should keep faces together
• Group photos benefit from content awareness

Consider Regular Mode Instead

⚠️ Abstract art — No clear boundaries to detect ⚠️ Uniform textures — Equal content everywhere ⚠️ Exact dimensions required — Smart mode may vary sizes ⚠️ Speed critical — Regular mode is faster

Comparison

Aspect	Smart Chunking	Regular Mode
Speed	Slower (analysis)	Faster
Chunk sizes	Variable	Fixed
Content awareness	Yes	No
Text protection	Yes	No
Best for	Varied content	Uniform content

Tips for Best Results

Improve Detection Accuracy

1. Use high-resolution sources — More pixels = better analysis
2. Increase contrast — Clear boundaries are easier to detect
3. Clean backgrounds — Reduce noise in whitespace areas

Handle Edge Cases

If Smart Chunking isn’t finding good splits:

1. Adjust sensitivity — Try higher or lower
2. Set minimum height — Prevent too-small chunks
3. Use overlap — Add pixels between chunks
4. Fall back to Regular — Sometimes fixed grids work better

Preview and Iterate

1. Process with Smart mode
2. Review each chunk boundary
3. Adjust settings if needed
4. Re-process until satisfied

Technical Details

Performance

• GPU Accelerated: Metal shaders analyze content
• Memory Efficient: Streaming processing for large images
• Cancellable: Stop analysis anytime

Accuracy Factors

Factor	Impact
Image resolution	Higher = better detection
Content contrast	Higher = clearer boundaries
Content variety	More variation = more options
Background uniformity	Cleaner = easier detection

Examples

Screenshot with Text

Original: 1080 × 4000px

Smart Chunking Result:
├── Chunk 1: Header + intro (1080 × 800px)
├── Chunk 2: Feature list (1080 × 1200px)
├── Chunk 3: Screenshots (1080 × 1100px)
└── Chunk 4: Footer (1080 × 900px)

Note: Each chunk contains complete sections

Webtoon Panel

Original: 800 × 10000px

Smart Chunking Result:
├── Chunks align with panel borders
├── Speech bubbles stay with their panels
└── No mid-action splits

Result: Natural reading experience when scrolled