How Object & watermark remover (Inpainting) Works Locally in Your Browser
Object & watermark remover (Inpainting) processes files and text entirely inside your browser tab using lama-fp16-pruned (Tier M). Zero bytes are uploaded to any server.
Remove unwanted objects, date stamps, or watermarks using fast-fourier convolution inpainting (LaMa-FP16-pruned ~38 MB on WebGPU) directly in your browser.
Step-by-Step Instructions
- Select or drop your image into the browser workspace.
- Configure processing parameters and preview the updated image in real time.
- Download your final processed image directly to your device with 0 server uploads.
Architecture & Standards Comparison
| Attribute | PrivacyFold Local Engine | Traditional Online Converters |
|---|---|---|
| Execution Runtime | lama-fp16-pruned (In-Browser RAM) | Remote Cloud Server Buckets |
| Standard Compliance | W3C WebAssembly Core 2.0 | Opaque Server Scripts |
| Network Footprint | 0 KB uploaded (Client-Side Only) | Full File Upload Over Internet |
| Offline Support | 100% Functional via Service Worker PWA | Fails Without Active Server Connection |
| Agent Integration | Deterministic JSON-RPC / MCP Tool Schema | Ad-Hoc REST Upload Form |
Frequently Asked Questions
Does Object & watermark remover (Inpainting) upload my files or text to a server?
No. Object & watermark remover (Inpainting) executes 100% inside your browser's local memory using lama-fp16-pruned. No data leaves your device.
Can AI agents or CLI scripts automate Object & watermark remover (Inpainting)?
Yes. Use Lane A schema /schemas/magic_eraser.json or open /tools/magic-eraser?mode=agent#payload with input[data-pf='input'] and window.__PRIVACYFOLD_RESULT__.