Magic Eraser

Why It Exists

Photos often contain unwanted objects like photobombers, trash cans, or distracting elements. Traditional object removal requires manual editing skills that most users lack.

How It Works

Magic Eraser is Google's AI-powered object removal tool that uses on-device Gemini Nano vision models to detect, select, and remove unwanted objects from photos. The system intelligently reconstructs the background behind the removed object, creating seamless results. It can remove photobombers, dust spots, wires, and other distractions. The tool also includes 'Motion Mode' for removing moving objects from busy backgrounds. All processing happens entirely on-device on supported Pixel phones, ensuring photo privacy.

Everyday Use Cases

• Removing photobombers
• Cleaning up backgrounds
• Erasing wires and poles
• Dust spot removal
• Motion blur cleanup
• Tourist removal from landmarks

User Workflow

• User opens a photo in Google Photos and selects Magic Eraser
• Gemini Nano scans the image for removable objects
• User selects objects to remove or uses auto-suggestions
• AI reconstructs the background seamlessly
• Result is saved as a new photo

AI Processing Flow

The photo is analyzed by on-device Gemini Nano vision models that detect objects and their boundaries. The system identifies elements that can be safely removed and suggests them to the user. Background reconstruction uses inpainting neural networks trained to match texture, lighting, and perspective. The inpainted region is blended seamlessly with surrounding pixels. All processing uses the Tensor NPU for fast results.

Inputs / Outputs

Inputs:

• Original photo
• Detected removable objects
• Background texture and lighting
• Depth information
• Scene context

Outputs:

• Edited photo with objects removed
• Seamlessly reconstructed background
• Preserved image quality
• Clean composition

Known Limitations

• Only available on Pixel devices with Tensor chips
• Large objects may leave artifacts
• Works best with simple backgrounds
• Cannot remove objects from videos
• Requires sufficient background detail for reconstruction

Unsupported Scenarios

• Video editing
• Professional photo retouching
• Non-Pixel devices
• Extremely crowded backgrounds
• Low-resolution images below 12MP

Performance Notes

Object detection in under 1 second on Tensor G3; Background reconstruction in 2 to 4 seconds; NPU acceleration for fast processing; Battery impact under 3 percent per use

Available On

Shows where this feature is available and how its AI processing works on each platform. Availability may vary by device.

Research Status

Confidence and verification reflect how complete documentation is. Fields may show Not assessed or Not yet verified while research is ongoing - this flags gaps, not product deficiencies.

Research Notes

Available on Pixel 7 and newer with Tensor chips; Introduced with Pixel 7 in 2022; Enhanced with Gemini Nano on Pixel 8 Pro 2023; Now supports motion object removal; 50+ million photos edited monthly

Details

Capability Mapping

This feature maps to canonical capability family: Image Generation.

Canonical capability: Image Generation.

• Image Generation - Creates novel imagery or emoji-like generated outputs from prompts.

Grounded in 3 device(s) on Gemini Nano.

Related Features

• AI Eraser - Cross-platform equivalent