Cropping and Framing in the AI Era
Cropping has always been a fundamental post-processing tool. AI has produced several automated cropping approaches that attempt to find the optimal composition within an image automatically. Understanding the difference between these tools helps you use them where they add genuine value.
Auto Crop: Simple Boundary Removal
Auto crop in its simplest form detects and removes unwanted borders: black bars, uneven edges from lens distortion correction, or white borders from scanned prints. It identifies the largest rectangular content area within an image and crops to that boundary. This is boundary detection, not compositional intelligence.
Most RAW editors including darktable and RawTherapee offer auto crop of this type. After applying lens distortion correction, the auto crop finds the largest usable rectangle and applies it automatically.
Smart Crop: AI-Driven Composition Optimisation
Smart crop attempts to find the most aesthetically pleasing crop of an image by analysing composition signals:
- Subject detection to ensure the primary subject remains in the cropped frame
- Rule-of-thirds placement to position subjects at compositionally strong positions
- Attention maps identifying regions of highest visual interest
- Aspect ratio fitting to find the best crop for a target ratio while preserving compositional integrity
Smart crop is genuinely useful for generating social media crops in multiple aspect ratios from landscape-format originals.
AI Composition Scoring vs Smart Crop
imagic includes composition as one of its five AI quality scoring dimensions alongside sharpness, exposure, noise, and detail. This scoring assesses the composition of the image as captured and is used to rank images within burst groups, not to suggest a new crop.
A low composition score in imagic indicates the as-captured framing may be weak, which is useful when choosing between similar frames in a burst sequence. Install imagic with pip install imagic to see composition scores in the Review step.
When to Use Each Tool
- Auto crop: After lens distortion correction, after straightening, when removing scan borders
- Smart crop: When generating social media crops in multiple aspect ratios from existing images
- AI composition scoring (imagic): During culling to identify the best-composed frames from a shoot or burst sequence
The Limits of AI Framing
AI framing tools optimise for learned compositional conventions and do not understand the narrative purpose of a specific crop or your creative intent. Always apply creative judgement on top of AI scoring output.