Quick Answer
An object remover erases unwanted things from a photo — strangers in the background, power lines, signs, watermarks — and fills the hole with pixels that match the surroundings, a technique called inpainting. Our free tool does this in your browser: upload the photo, paint a rough mask over the object, and the LaMa inpainting neural network reconstructs the background underneath. The AI model downloads once and then runs locally, so your photo is never uploaded to any server.
How It Works
Upload a PNG, JPEG, or WebP photo — it is processed locally and never uploaded
Paint a rough mask over the unwanted object with the red brush; adjust brush size, erase mistakes, undo strokes, or clear the mask entirely
Click Remove to run LaMa AI inpainting locally (model downloads once, then is cached) — or use the instant Fast fill engine that needs no model at all
Compare before and after, then download the clean PNG — free, no watermarks, no sign-up
Key Facts
- Paint a rough brush mask over any unwanted object — no precise selection needed
- LaMa AI inpainting runs locally via onnxruntime-web; nothing is uploaded
- Adjustable brush size (8–120 px), eraser mode, undo, and clear-mask controls
- Two verified open-source models: a fast 60 MB int8 build and a full-quality 200 MB Big-LaMa
- Instant "Fast fill" fallback works even with no AI model (deterministic push-pull inpainting)
- Before/after comparison and one-click PNG download at full working resolution
- Free with no sign-up, no watermarks, and no usage limits
Frequently Asked Questions
How do I remove an object from a photo for free?
Upload your photo, paint over the object you want gone with the red brush (adjust the brush size for big or small objects), then click Remove. The AI inpainting model reconstructs the background where the object stood — typically in under 10 seconds. Click Download PNG to save the clean image. It is completely free with no sign-up and no watermarks.
Is my photo uploaded anywhere?
No. The object remover runs entirely in your browser. The only network traffic is a one-time download of the open-source AI model files from a public CDN (Hugging Face / jsDelivr) — your photo itself never leaves your device, is never transmitted, and is never stored. After the model is cached you can even go offline and keep editing.
Is this object remover really free? What are the limits?
Yes, completely free — no account, no credits, no watermarks, and no daily limits. Images are processed at up to 1024 px on the longest side for reliable AI results (larger uploads are downscaled automatically). The AI model is 60 MB (fast) or 200 MB (best quality) on first use and is cached by your browser afterwards. The instant Fast fill mode has no download at all.
Why does the first removal take longer?
The first time you use AI removal, your browser downloads the LaMa inpainting model (about 60 MB in fast mode, 200 MB in best mode) and caches it. Every later run starts almost immediately and only spends time on inference. If you do not want any download, use the Fast fill engine — it is instant and works offline.
What kinds of objects can it remove — and when does it struggle?
It works best on objects surrounded by predictable background: sky, water, grass, walls, pavement, blur. Strangers, wires, signs, logos, small watermarks, and blemishes are ideal. It can struggle when the masked area covers complex repeating texture or a large portion of the image — in that case mask a smaller region at a time, or switch between the fast and best-quality models. The mask is dilated a few pixels automatically to catch soft edges.