Magic Eraser (AI Object Remover)
Powered by LaMa, an open-source AI inpainting model, running entirely in your browser via WebAssembly — no file is ever uploaded anywhere. The first use downloads the AI model (~200MB, one-time, cached by your browser afterward).
Choose an image (JPG, PNG, or WEBP)Brush over the object you want removed (shown in red). Hold Space to pan, scroll to zoom, Ctrl+Z / Ctrl+Y to undo/redo. Works with touch, mouse, or stylus.
About the Magic Eraser (AI Object Remover)
Photobombers, stray objects, power lines, blemishes — Magic Eraser removes them properly. Brush over what you want gone, and an open-source AI inpainting model (LaMa) reconstructs the area based on the surrounding image content, running entirely in your browser. This is genuine generative inpainting, the same category of technology behind tools like Adobe's Generative Fill, not a blur, clone-stamp, or flat color patch.
Features
- Adjustable brush size and soft edge for precise selections
- Zoom, pan, and fit-to-screen for detailed work on large photos
- Undo, redo, clear selection, and reset image
- Before/after comparison with a draggable split slider
- Download as PNG or JPG when you're done
How to Use
Upload a photo, adjust the brush size, and paint over the object you want removed (shown highlighted in red). Tap Remove Object — the AI model downloads once on first use, then processes your selection. Compare the before/after with the slider, and download your result.
Benefits
- Real AI reconstruction of the erased area, not a copied patch or blur
- Nothing is uploaded — the AI model and all processing run locally in your browser
- Works alongside ToolFlight's other image tools — remove an object here, then crop, watermark, or compress the result
Frequently Asked Questions
Is this really AI, or just a blur/clone tool?
Genuinely AI: this uses LaMa (Large Mask Inpainting), an open-source neural network trained specifically for this task, run through ONNX Runtime Web. It reconstructs the erased area based on the surrounding image content — it does not blur, clone nearby pixels, or fill with a flat color.
Why does the first use take a while?
The AI model is roughly 200MB and downloads once on first use. Your browser caches it afterward, so later uses on the same device are much faster.
Does this work on very large or very detailed objects?
It works best on small to medium objects with a reasonably plain or repeating surrounding background — like most inpainting tools, very large or highly detailed removals are more likely to show visible artifacts.
Is my photo uploaded anywhere?
No — both the AI model and all image processing run entirely inside your browser. Your photo is never sent to a server.