AI Watermark Remover
Brush over watermarks or unwanted objects — AI fills the area naturally. All processing runs on your device.
Downloading AI model (28 MB)...
Drop an image here or click to browse
Supports JPG, PNG, WebP. AI processing is 100% local.
About this tool
This tool uses MI-GAN, a deep learning model designed for mobile devices, to intelligently fill areas you mark. The model analyzes the surrounding image content (textures, edges, patterns) and reconstructs what should be behind the watermark or object.
The AI model (~28 MB) downloads once and is cached in your browser for future use. All inference runs locally via ONNX Runtime Web with WebAssembly. Nothing is ever uploaded.
How to use
- Upload an image — drag & drop or click to browse.
- Paint over the watermark — use the brush to mark the area you want removed. The red overlay shows what's selected.
- Click "Inpaint" — the AI model processes the image. This takes ~1–3 seconds depending on your device.
- Review & download — toggle between Result / Original / Mask views to compare. Download the cleaned image.
FAQ
Does it really run AI in the browser?
Yes. The MI-GAN inpainting model runs via ONNX Runtime Web + WebAssembly directly in your browser. The model (~28 MB) downloads once and is cached in IndexedDB.
How good is the inpainting quality?
MI-GAN works well on smooth backgrounds, gradients, and simple textures. It may struggle with highly complex patterns. Best results come from marking only the watermark itself (not a huge area).
Is my image uploaded anywhere?
No. Everything runs locally. The model runs in your browser via WebAssembly. Your image never leaves your device.