📡 LANStream

Low-latency screen streaming over your local network Windows → Linux — GPU-accelerated capture, hardware H.264/HEVC encoding, OpenGL rendering

---

✨ What is LANStream?

LANStream streams your Windows desktop to a Linux machine on the same network with 10–20 ms glass-to-glass latency. It uses:

• dxcam for near-zero-CPU DirectX GPU capture on Windows
• ffmpeg NVENC for hardware H.264/HEVC encoding (falls back to libx264 if no NVIDIA GPU)
• RTP/UDP for efficient, low-overhead video transport
• PyAV for hardware-capable decode on the Linux client
• QOpenGLWidget + GLSL for GPU-accelerated display with a sub-millisecond texture upload

---

🏗️ Architecture

Windows (server.py)                         Linux (client.py)
────────────────────────────────────────    ──────────────────────────────────────
dxcam  (DirectX 11 GPU capture, ~0 CPU)
  │  raw BGRA frames via pipe
ffmpeg NVENC  (H.264 / HEVC hardware encode)
  │  RTP/UDP stream over LAN
                                       ──►  PyAV  (RTP demux + H.264/HEVC decode)
                                              │  decoded RGB frames
                                            QOpenGLWidget  (GLSL texture → screen)
                                              │
                                            KDE Plasma / Wayland / X11

A separate TCP control channel carries authentication, client registration, and keepalive messages using newline-delimited JSON — fully compatible with v1 clients.

---

⚡ v1 → v2 Improvements

Component	v1	v2
Capture	dxcam / mss	dxcam / mss (same)
Encode	cv2 JPEG + custom UDP	ffmpeg NVENC H.264/HEVC
Transport	Raw UDP chunked frames	Standard RTP
Decode	cv2.imdecode JPEG	PyAV H.264/HEVC
Display	QPainter QPixmap	QOpenGLWidget + GLSL shader
Server CPU	Medium (JPEG encode)	Near zero (NVENC on GPU)
Client CPU	Medium (cv2 decode)	Low (libavcodec + GL upload)
Latency	~30–60 ms	~10–20 ms

---

📋 Requirements

🪟 Server — Windows

Requirement	Details
Python	3.10 or newer
ffmpeg	In PATH, with NVENC support — download here (recommend BtbN/FFmpeg-Builds)
NVIDIA GPU	Required for NVENC; auto-falls back to libx264 otherwise

pip install -r requirements_server.txt

No NVIDIA GPU? No problem — the server detects codec availability at startup and automatically falls back to libx264 software encoding. Same RTP output, just slightly higher CPU usage.

---

🐧 Client — Linux (Arch / Ubuntu / Debian)

Arch Linux:

sudo pacman -S ffmpeg qt6-base python-pyqt6 python-numpy
pip install av PyOpenGL

Ubuntu / Debian:

sudo apt install ffmpeg python3-pyqt6 python3-numpy python3-opengl
pip install av

Or use the requirements file:

pip install -r requirements_client.txt

---

🚀 Quick Start

1. Start the server (Windows)

python server.py

• Select your monitor, codec, bitrate, and FPS in the GUI
• Click ▶ Start Streaming
• The server announces itself via UDP multicast — Linux clients discover it automatically

2. Start the client (Linux)

python client.py

• Click ⚡ Connect
• If auto-discovery found the server, it appears in the dropdown
• Otherwise, enter the server IP manually
• The server assigns an RTP port automatically per client

---

⌨️ Keyboard Shortcuts (Client)

Key	Action
F11	Toggle fullscreen
Escape	Exit fullscreen
Ctrl+H	Toggle HUD overlay
Ctrl+N	Open connect dialog
Ctrl+D	Disconnect
Ctrl+Q	Quit

---

🎛️ Codec & Bitrate Guide

Scenario	Codec	Bitrate	Notes
Gaming / fast motion	h264_nvenc	8000k–16000k	p4 preset
Desktop / productivity	h264_nvenc	4000k–6000k	p4 preset
Slow LAN / WiFi	h264_nvenc	2000k–4000k	p3 preset
No NVIDIA GPU	libx264	4000k–8000k	ultrafast auto-selected
Better quality (HEVC)	hevc_nvenc	4000k–8000k	Requires PyAV with HEVC

NVENC presets: p1 = fastest (lowest quality) → p7 = slowest (best quality). p4 is the recommended balance for 60 FPS gaming.

---

🔒 Firewall Ports

Port	Protocol	Direction	Purpose
9876	TCP	client → server	Control channel (auth, registration, keepalive)
9877	UDP	multicast	Server auto-discovery
5004+	UDP	server → client	RTP video stream (one port per client)

RTP ports are assigned from 5004 upward — each connected client gets its own port.

---

🛠️ Troubleshooting

ffmpeg: command not found

Add ffmpeg to your Windows PATH. Verify it works by running ffmpeg -version in Command Prompt.

NVENC not available

Run ffmpeg -encoders | findstr nvenc — if the output is empty, your ffmpeg build lacks NVENC support or you have no NVIDIA GPU. The server will automatically fall back to libx264.

PyAV not found

pip install av

PyAV bundles its own ffmpeg libraries on Linux, so you don't need a separate ffmpeg install just for the client.

PyOpenGL not found

pip install PyOpenGL

If PyOpenGL is missing, the client automatically falls back to a QPainter/QPixmap renderer — slightly higher CPU usage but identical output.

Black screen / no video

• Verify your firewall allows UDP on the RTP port shown in the client's status bar
• Confirm the RTP listen port in the connect dialog matches what the server assigned
• Temporarily disable the Windows Firewall to rule it out during testing

High latency

• Use a wired LAN connection — WiFi adds 5–20 ms jitter
• Increase the bitrate so the encoder doesn't drop frames
• Use NVENC preset p1 or p2 for minimum encode latency

---

📁 Project Structure

LANStream/
├── server.py                  # Windows capture & streaming server (GUI)
├── client.py                  # Linux RTP receiver & OpenGL viewer (GUI)
├── requirements_server.txt    # Python dependencies for the server
├── requirements_client.txt    # Python dependencies for the client
└── README.md

---

🤝 Contributing

Pull requests are welcome! For major changes, please open an issue first to discuss what you'd like to change.

1. Fork the repository
2. Create your feature branch: git checkout -b feature/my-feature
3. Commit your changes: git commit -m 'Add some feature'
4. Push to the branch: git push origin feature/my-feature
5. Open a Pull Request

---

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

---

Made with ❤️ for low-latency LAN streaming