zheji

Project Overview

Zheji is a Chinese-language local image privacy tool for covering sensitive areas such as phone numbers, addresses, avatars and order IDs before sharing a screenshot, cropping as needed, adding a purpose watermark, and exporting a new PNG. Images are processed only in page memory with no uploads, signup or AI API, and the production build works offline once loaded. It is built with React and TypeScript, and the core processing has no networked services.

Keep a little privacy before sharing, all processed locally.

Key Features

• Covering: opaque rectangular covers with reverse dragging, coordinate fine-tuning, multiple regions and undo/redo
• Cropping: select a region to keep and discard the rest on export
• Purpose watermark: repeating diagonal watermark with custom text and opacity
• Preview and export: shows the exact PNG to be exported before download, not an approximation
• Metadata cleaning: output keeps only the pixel-decoding chunks without original EXIF or other metadata
• Local and offline: works offline after the production build has loaded once

Target Audience

• Users who frequently share chat, order and logistics screenshots
• Privacy-conscious users who want to cover sensitive info before sharing
• Users who do not want images uploaded to the cloud

Quick Start

1. Open the online version at fuzzylogic112.github.io/zheji or self-host it
2. Choose an image or try the fictional example
3. Drag out rectangular covers, then crop or enable the watermark as needed
4. Click "Review and export", check the final preview and download the PNG
5. Recheck the downloaded file and share only that new image

Core Advantages

• Local processing: images never leave the browser, with no upload, signup or AI API
• What you see is what you get: the exported file is exactly the previewed PNG
• Offline capable: keeps working without a network after the first load

System Requirements

Item	Requirement
Input formats	PNG, 8-bit baseline/progressive JPEG
Size limits	Source ≤20 MiB, total pixels ≤20 million
Runtime	Modern browser (Chromium verified)

Notes

• The tool does not automatically find missed information; uncovered text, QR codes and more can still leak privacy
• Solid-color covers only hide the selected areas and do not certify the whole image as anonymized
• Output clears only non-pixel metadata and does not detect steganographic content inside pixels