README.md

---
title: Watermark Remover
emoji: 🖼️
colorFrom: indigo
colorTo: purple
sdk: docker
app_port: 7860
pinned: false
---

# Watermark Remover

Remove watermarks from images and videos using AI inpainting.

## Features

- **Manual masking** — Brush, Box, or Eraser tools
- **SAM ✦** — Click on the watermark, AI segments it precisely (Segment Anything Model)
- **Auto Detect** — EasyOCR detects text watermarks automatically
- **Three inpainting engines:**
  - OpenCV — instant, solid quality
  - LaMa — ML-based, better quality
  - Stable Diffusion — near-invisible results (~20s on GPU)
- **Before/After slider** — drag to compare original vs result
- **Video support** — Fast mode (stamp first frame) or Thorough (per-frame inpainting)
- **Audio preserved** — original audio track remuxed into output

## Local development

```bash
pip install -r requirements.txt
pip install ffmpeg  # or: brew install ffmpeg
uvicorn app:app --port 8000
# → open http://localhost:8000
```

## File storage

Uploads and results are stored in `/tmp/wm_tool/` — cleaned up on OS restart.