(Formerly AIShare-Grabber)
Transfer photos directly from your Android phone to your PC over high-speed local Wi-Fi.
Then search and organize your entire photo library using 100% on-device AI.
🚫 No Cloud Uploads • 🔌 No USB Cables • 💳 No Subscriptions • 🔒 100% Private
👉 官方门户与全平台下载直达：https://novamindlab.github.io/AIShare-Grabber/ 👈
（支持全平台客户端一键下载、Mac 芯片架构智能检测与免安装 WebShare 在线体验）
🎬 Demo • 💡 Real Scenarios • ⚖️ Comparison • ✨ Features • 🚀 Quick Start • 🧠 Under the Hood • 🇨🇳 中文说明
┌─────────────────┐ Local Wi-Fi / P2P ┌─────────────────┐
│ Android 📱 │ ──────────────────────────────> │ PC 🖥️ │
│ (Scan QR Code) │ (Zero Cloud • 80+ MB/s) │ (Local Storage) │
└─────────────────┘ └────────┬────────┘
│
▼
┌─────────────────┐
│ Local AI 🧠 │
│ (100% Offline) │
└────────┬────────┘
│
▼
🔍 "dog playing on grass"
🔍 "sunset at the beach"
🔍 "receipts from last month"
- Instant Wireless Pairing: Open the Android companion app, scan the QR code on your PC screen—connected in seconds.
- Lightning P2P Transfer: Your photos stream directly over your local Wi-Fi at gigabit speeds (80+ MB/s). No USB cables, no mobile data usage.
- On-Device AI Indexing: PC indexes and classifies your photos completely locally using embedded vision models.
- Natural Language Search: Type what you remember in everyday language to find any photo instantly.
Your photos never route through external servers. Mobile and desktop discover each other offline via Bluetooth Low Energy (BLE) or LAN broadcast, negotiating a direct WebRTC P2P DataChannel over local Wi-Fi:
🎥 Interactive Web Demo & Video: Watch the full 1080P real-world walkthrough on our Official Website or try the Zero-Install WebShare.
“Transfer 1,000 vacation photos from Android to PC without hunting for a USB-C cable.”
- No driver installation, no ADB configuration, and no cable swapping.
- Tap Sync, and high-resolution thumbnails and photos stream straight into your organized PC folders at local Wi-Fi line rate.
- Backpressure flow control ensures butter-smooth transfers without choking your phone or PC.
“Where was that photo of my golden retriever in the snow?”
Instead of scrolling through thousands of file names like IMG_20240827_103211.jpg, just search:
"golden retriever running in the snow""sunset at the beach""tax invoices and receipts""coffee latte art on wooden table"
The embedded MobileCLIP multimodal AI maps both your query and your pictures into a shared 512-dimensional vector space completely offline.
“Group all photos of family members automatically into dedicated albums.”
- Cascaded on-device face pipeline (SCRFD 500M detection + MobileFaceNet 512-D ArcFace).
- Groups portraits by individual identity using vectorized clustering—all calculated on your CPU/GPU without sharing biometric profiles to any third party.
“Keep personal memories, family moments, and sensitive documents off the cloud.”
- Zero Cloud Upload: Everything stays on your physical devices.
- Zero Telemetry: No tracking, no user profiling, no subscriptions.
- Offline Capable: Works even without an internet connection (e.g., local router or phone mobile hotspot).
| Capability / Feature | ✨ ShareCLIP | ☁️ Cloud Albums (Google / iCloud) | 🔌 USB Cable Transfer |
|---|---|---|---|
| P2P Wireless Transfer | ✅ Yes (Gigabit Wi-Fi Direct) | ❌ No (Routes to remote server) | ❌ No (Wired only) |
| Zero Cloud Upload | ✅ 100% Private & Local | ❌ Uploaded to 3rd-party servers | ✅ Local only |
| Natural Language AI Search | ✅ Local AI (100% Offline) | ❌ No search capability | |
| Facial Recognition & Albums | ✅ On-Device (WASM / SIMD) | ❌ No | |
| Requires Physical Cable | ❌ No (Wireless QR link) | ❌ No | |
| Storage & Subscription Cost | 🆓 Free & Open Source | 💳 Monthly storage fee ($2~$10/mo) | 🆓 Free |
| Internet / Cellular Data Usage | ⚡ 0 MB (LAN traffic only) | 📶 Consumes broadband/data plan | ⚡ 0 MB |
| Cross-Device Deletion & Cleanup | ✅ Synchronized cleanup | ❌ Manual file browsing |
- ⚡ Ultra-Fast Local P2P Sync: Seamless discovery over BLE GATT / UDP broadcast, automatically establishing encrypted WebRTC DataChannels for gigabit transfers.
- 🔍 Multimodal Semantic Image Search: 512-D neural embeddings enable natural text-to-photo queries without manual tagging.
- 👥 Biometric Face Recognition & People Albums: Fast face detection and clustering grouping family, friends, and portraits in seconds.
- 🧹 Intelligent Duplicate & Burst Shot Cleanup: Cosine similarity leader-clustering locates duplicate shots and lets you free disk space on both PC and phone with one click.
- 🗺️ EXIF Travel Footprint Map: Extracts GPS coordinates from camera EXIF tags, plotting your life memories on an interactive clustered map.
- 🖼️ 4K Immersive Lightbox: Instant thumbnail browsing with on-demand 4K RAW image streaming from your phone when connected.
- 🌐 20+ Global Languages: Fully internationalized UI (English, 简体中文, 繁體中文, 日本語, 한국어, Español, Français, Deutsch, etc.).
📱 One-Tap Phone Connection & Storage Status
|
⚡ High-Speed Transfer & AI Synchronization
|
🧹 Smart Duplicate & Similar Image Clustering
|
🔍 Local Natural Language Query & Detection
|
🌐 Official Website (官方网站): https://novamindlab.github.io/AIShare-Grabber/
(Visit the official portal for auto-detected hardware downloads, Mac M-Series / Intel chip selector, and interactive WebShare.)
- PC & Mac Desktop: Download the Windows installer (
.exe) or macOS installer (.dmgfor Apple Silicon & Intel) from the Official Website or GitHub Releases. - Android Companion: Download and install the companion APK (
.apk) on your phone. - Connect: Launch both apps on the same Wi-Fi network, scan the QR code on your PC screen, and start syncing!
No installation on PC? Try WebShare Online directly in Google Chrome or Microsoft Edge.
AIShare-Grabber/
├── 📱 android/ # Flutter Android Companion App (BLE GATT + WebRTC DataChannel)
├── 🖥️ cp_clip/ # Desktop Host App (Electron 30, Vue 3, ONNX Runtime, SQLite, Sharp)
│ ├── src/workers/ # Worker Thread Pool (Search, SIMD WASM clustering, ONNX Inference)
│ └── models/ # Quantized MobileCLIP-S0 & Buffalo_SC ONNX models
├── 🌐 web/ # Official Portal & Interactive Demo Website
├── ⚡ webshare/ # Pure Web Client (WebGPU + WebRTC P2P + IndexedDB)
└── 📖 wiki/ # Comprehensive Architecture & Protocol Documentation
# Prerequisites: Flutter SDK >= 3.11.1 & Android SDK
cd android
flutter pub get
flutter run --release# Prerequisites: Node.js >= 20, Python >= 3.10
cd cp_clip
npm install
# Run in Development Mode
npm run dev
# Package as Windows Executable
npm run distcd web
npm install
npm run devFor system architects, AI researchers, and contributors who want to dive deeper into the implementation:
flowchart LR
subgraph Mobile [Android Companion / WebShare]
A[MediaStore / PhotoManager] -->|15-Way Concurrent Extraction| B[400x400 Thumbnail Stream]
B -->|Chunked 16KB-32KB Binary Packets| C[WebRTC DataChannel]
end
subgraph Desktop [PC Host Electron]
C -->|1-2ms Rapid Disk Write| D[Local Isolated Storage]
D -->|Task Queue Dispatch| E[ONNX Worker Pool]
subgraph AI_Engine [On-Device AI Acceleration]
E -->|MobileCLIP-S0 256x256| F[512-D Image Embedding]
E -->|SCRFD + MobileFaceNet| G[Face Biometric 512-D]
F & G -->|Lock-Free Single Writer| H[(SharedArrayBuffer 40MB-200MB)]
end
H -->|Zero-Copy SIMD Dot Product| I[Sub-Millisecond Cosine Search]
I --> J[Vue 3 60fps Virtual Timeline]
end
- Lock-Free SharedArrayBuffer RAM Sharing: Eliminates IPC serialization overhead by mapping 512-D embeddings directly into shared physical memory (40MB~200MB tiering). Vector clustering of 5,000 photos finishes in ~120ms.
- 16-Byte Binary Protocol Framing: Fixed header protocol (
fileId,chunkIndex,totalChunks,payloadLength) handles high-concurrency streaming with backpressure control. - DirectML GPU ➔ CPU Auto-Fallback: Ensures flawless operation across diverse Windows hardware configurations (discrete GPUs, AMD APUs, Intel integrated graphics).
- Deep Architectural Specifications: Read the complete design in our Project Wiki:
- AI Architecture & Zero-Copy SAB Deep Dive
- WebRTC Channel Protocol & 16-Byte Headers
- Face Recognition & Same-Photo Exclusion Clustering
- BLE GATT Signaling & Fallback Architecture
ShareCLIP (原 AIShare-Grabber) 是一款完全面向用户、注重隐私与极速体验的开源跨端照片同步与本地 AI 相册管理软件。
Tip
🌐 官方门户与全平台下载直达：https://novamindlab.github.io/AIShare-Grabber/
官网提供各平台安装包一键下载（支持 Windows、macOS M系列 / Intel 双架构智能适配、Android 及免安装 WebShare 网页体验）。
日常生活中，我们经常面临两大痛点：
- 传图繁琐：手机里几百上千张旅行照片想备份到电脑，要么费劲翻找 Type-C 数据线，要么被各种网盘上传下载限速、消耗大量手机流量；
- 搜图如大海捞针：电脑上存了几万张照片，全是以
IMG_...命名的文件，想找某一张照片只能一张张肉眼翻看。
ShareCLIP 彻底解决了这两个问题：
- 手机 ➔ 电脑高速直传：手机扫描电脑屏幕二维码，近场蓝牙秒速配对并打通局域网 WebRTC P2P 直连通道，以 80+ MB/s 千兆速度无线极速同步，不经过任何公网服务器，不消耗手机移动流量，不需要 USB 数据线。
- 100% 离线本地 AI 相册：内置轻量级视觉模型（MobileCLIP），直接用大白话搜图（如 “海边日落”、“在草地奔跑的金毛”、“去年夏天的聚会发票”），毫秒级呈现结果。人脸聚类、连拍去重、旅行足迹全部在本地电脑运行，照片和生物特征绝不上云。
- ⚡ 极速无网互传：告别数据线，局域网千兆直接互通；
- 🔒 100% 本地隐私安全：零云端、零回传、零跟踪，真正的端侧自主权；
- 🧠 自然语言以文搜图：告别冰冷文件名，输入想找的画面自然呈现；
- 👥 本地人脸聚类：自动将家人、朋友分类成专属人物相册；
- 🧹 连拍与相似图清理：一键定位废片，同步释放手机与电脑存储；
- 🆓 完全免费与开源：基于 MIT 开源协议，无订阅套路。
- Privacy Guarantee: ShareCLIP is strictly engineered around a Zero-Cloud & Zero-Telemetry architecture. Your original photos, location EXIF data, and face biometric vectors will never leave your local devices.
- License: Released under the MIT License.
Developed with ❤️ by the NovaMindLab team.