QR Transfer

Offline file transfer via QR codes. One single HTML file — no server, no cloud, no tracking.

Try it live | Download HTML

How it works

1. Open index.html on both devices (sender and receiver)
2. Sender: Select a file → QR codes are generated and displayed in sequence
3. Receiver: Start camera → point at sender's screen → chunks are scanned automatically
4. When all chunks are received, the file is assembled with CRC32 integrity check and offered for download

Features

• Single HTML file — everything bundled, works offline
• No server required — runs entirely in the browser
• CRC32 integrity check — detects corrupted transfers
• Gzip compression — reduces transfer size
• Configurable chunk size and speed — tune for your setup
• Visual chunk grid — see transfer progress in real-time
• Drag & drop file selection
• Responsive — works on desktop and mobile

Quick Start

1. Online: Open the live demo on both devices
2. Offline: Download index.html and open it in any modern browser

Default settings: 300 B chunk size, 100 ms speed.

Development

npm install
npm run dev      # Start dev server
npm test         # Run unit tests
npm run build    # Build single HTML file to dist/

Protocol

Files are compressed (gzip level 9), base64-encoded, and split into chunks. Each chunk becomes a QR code containing a JSON packet:

{
  "v": 1,           // protocol version
  "i": 0,           // chunk index
  "t": 50,          // total chunks
  "n": "file.pdf",  // filename (chunk 0 only)
  "s": 123456,      // file size in bytes (chunk 0 only)
  "h": "a1b2c3d4",  // CRC32 hex of original file (chunk 0 only)
  "d": "base64..."  // chunk data
}

The receiver collects chunks in any order, reassembles, decompresses, and verifies the CRC32 checksum.

Smart Transfer (Bidirectional Feedback)

Smart Transfer creates a bidirectional channel where the receiver tells the sender which chunks are still missing via audio tones (like a modem). The sender then skips already-received chunks, dramatically speeding up the transfer.

How it works

• Data flows via QR codes (sender screen → receiver camera) — fast, visual
• Feedback flows via audio FSK tones (receiver speaker → sender microphone) — omnidirectional, no line-of-sight needed

  Device A (Smart Send)           Device B (Smart Receive)
  ┌──────────────────┐            ┌──────────────────┐
  │                  │            │                  │
  │   [DATA QR]      │ ── eye ──► │   [CAMERA]       │
  │   chunks 3,7,12  │            │   scanning QR    │
  │                  │            │                  │
  │   [MICROPHONE]   │ ◄─ ear ── │   [SPEAKER]      │
  │   listening      │    audio   │   FSK tones      │
  │                  │            │   "got 1,2,4,5"  │
  └──────────────────┘            └──────────────────┘

The phone can lie flat on a table — audio works in all directions, no camera positioning needed.

Audio Feedback Protocol (FSK Modem)

The feedback is encoded as FSK (Frequency Shift Keying) audio:

• 1200 Hz = bit "0", 2400 Hz = bit "1" (Bell 202 standard)
• 300 baud — each bit lasts ~3.3ms
• Frame format: [preamble 0xAA 0xAA] [sync 0xD5] [length] [bitfield] [CRC-8]
• Bitfield: 1 bit per chunk (received=1, missing=0), raw bytes
• CRC-8/CCITT: integrity check, invalid frames are silently discarded

For 100 chunks: 13 bytes bitfield → 144 bits total → 0.48 seconds per feedback cycle.

The receiver plays the feedback audio in a continuous loop. The sender's microphone picks it up, decodes via Goertzel algorithm, and adapts its QR playlist to only show missing chunks.

Recommended file sizes

Size	Transfer time (est.)	Experience
< 10 KB	seconds	Instant
10-100 KB	1-5 min	Comfortable
100 KB - 1 MB	5-30 min	Feasible
> 1 MB	30+ min	Use only as last resort

Tech Stack

• TypeScript
• Vite + vite-plugin-singlefile
• qrcode — QR generation
• jsQR — QR scanning
• pako — gzip compression
• Vitest — unit testing

License

MIT