# rtsp-server

**Repository Path**: xhoobin/rtsp-server

## Basic Information

- **Project Name**: rtsp-server
- **Description**: 海康摄像头AI推理后，零延迟的 AI 视频流 RTSP 服务器
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 60
- **Created**: 2026-04-25
- **Last Updated**: 2026-04-25

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# RTSP Server - Ultra-Low Latency AI Video Streaming

[中文](README.md) | **English**

> **Say goodbye to lag!** Stream your AI detection results with near-zero latency!

![Demo](docs/rtsp服务器0延迟效果.gif)

A standalone RTSP server module based on live555, supporting NVENC hardware encoding for ultra-low latency video streaming. **Supports auto-registration to ZLMediaKit for WebRTC playback.**

## Features

- **NVENC Hardware Encoding**: Uses NVIDIA GPU for H.264 encoding with <5ms latency
- **Ultra-Low Latency**: Single frame buffer queue for minimal delay
- **Auto ZLM Registration**: Automatically registers to ZLMediaKit on startup
- **Command Line Arguments**: Flexible configuration for video source, port, ZLM settings
- **Auto Reconnect**: Automatically reconnects when video source disconnects
- **Standalone Module**: Can be used independently from the main project

## Requirements

- Windows 10/11
- Visual Studio 2022 (Build Tools)
- CMake 3.15+
- vcpkg with packages: `opencv4`, `glog`, `ffmpeg`
- NVIDIA GPU with NVENC support (optional, falls back to software encoding)
- ZLMediaKit (for WebRTC forwarding)

## Quick Start

### 1. Build

```batch
cd rtsp_server
build.bat
```

### 2. Run

```batch
cd build\Release
RTSPServer.exe rtsp://admin:password@192.168.1.64:554/stream --zlm-secret YOUR_ZLM_SECRET
```

### 3. Play

Open WebRTC player in browser, or use VLC to play `rtsp://localhost:8554/ai_stream`

## Command Line Arguments

```
RTSPServer.exe <rtsp_source> [options]

Options:
  --port <port>           RTSP server port (default: 8554)
  --stream <name>         Stream name (default: ai_stream)
  --zlm-host <host>       ZLM host (default: localhost)
  --zlm-port <port>       ZLM HTTP port (default: 8088)
  --zlm-secret <secret>   ZLM API secret (required for auto-register)
  --no-zlm                Disable auto-register to ZLM
  --no-display            Disable OpenCV window display
  --help                  Show help message
```

## Usage Examples

### Basic Usage (Auto-register to ZLM)

```batch
RTSPServer.exe rtsp://admin:password@192.168.1.64:554/Streaming/Channels/102 --zlm-secret hXSs7WNDSONHwG7vdhWL2qmZMgnDL2D6
```

### Custom Port and Stream Name

```batch
RTSPServer.exe rtsp://admin:password@192.168.1.64:554/stream --port 8555 --stream camera1 --zlm-secret abc123
```

### Connect to Remote ZLM Server

```batch
RTSPServer.exe rtsp://admin:password@192.168.1.64:554/stream --zlm-host 192.168.1.100 --zlm-port 8088 --zlm-secret abc123
```

### RTSP Server Only (No ZLM)

```batch
RTSPServer.exe rtsp://admin:password@192.168.1.64:554/stream --no-zlm
```

### Background Mode (No Display)

```batch
RTSPServer.exe rtsp://admin:password@192.168.1.64:554/stream --zlm-secret abc123 --no-display
```

## Workflow

```
┌─────────────────┐     ┌──────────────────┐     ┌─────────────────┐     ┌─────────────────┐
│  Camera/Source  │────>│  RTSPServer.exe  │────>│   ZLMediaKit    │────>│  Browser/Player │
│  (RTSP/File)    │     │  (NVENC Encode)  │     │  (WebRTC Relay) │     │  (WebRTC Play)  │
└─────────────────┘     └──────────────────┘     └─────────────────┘     └─────────────────┘
                               │                        │
                               │ Auto Register          │ WebRTC
                               │ addStreamProxy         │
                               ▼                        ▼
                        rtsp://localhost:8554    http://localhost:8088
```

### 📸 Demo: Original Stream vs AI Stream

![Zero Latency Demo](docs/rtsp服务器0延迟效果.gif)

![Original vs AI Stream](docs/原始视频流与AI视频流对比.png)

> Left: Original camera feed. Right: AI detection output. Nearly synchronized!

## Playback

### WebRTC (Recommended, Lowest Latency)

After startup, the program outputs the WebRTC URL:
```
[INFO] WebRTC URL: http://localhost:8088/index/api/webrtc?app=live&stream=ai_stream
```

### VLC Player

```
Media -> Open Network Stream -> rtsp://localhost:8554/ai_stream
```

### FFplay (Low Latency Mode)

```batch
ffplay -fflags nobuffer -flags low_delay rtsp://localhost:8554/ai_stream
```

## Performance

| Metric | Value |
|--------|-------|
| Encoding Latency | <5ms (NVENC) |
| End-to-End Latency | 50-100ms (WebRTC) |
| Max Resolution | 4K |
| Default Bitrate | 4Mbps |
| GOP Size | 30 frames |

## Troubleshooting

### ZLM Registration Failed
1. Ensure ZLMediaKit is running
2. Check `--zlm-secret` is correct
3. Check `--zlm-port` is correct (default 8088)

### Video Source Not Found
1. Verify RTSP URL is correct
2. Ensure camera is online and accessible
3. Test RTSP URL with VLC first

### NVENC Not Available
The program automatically falls back to x264 software encoding.

### CMake Build Failed
Use English-only paths. CMake 4.x has issues with non-ASCII paths.

## Directory Structure

```
rtsp_server/
├── RTSPServer.h          # Header file
├── RTSPServer.cpp        # Implementation
├── example_main.cpp      # Main program with CLI support
├── CMakeLists.txt        # CMake build file
├── build.bat             # Build script
├── README.md             # Chinese documentation (default)
├── README_EN.md          # English documentation
├── 3rdparty/
│   └── live555/          # live555 library source
└── build/
    └── Release/
        ├── RTSPServer.exe  # Executable
        └── *.dll           # Required DLLs
```

## API Reference (For Development)

```cpp
class RTSPServerWrapper {
public:
    bool init(int port, const std::string& streamName, int width, int height, int fps);
    bool start();
    void stop();
    bool pushFrame(const cv::Mat& frame);
    std::string getRtspUrl() const;
    bool isRunning() const;
};
```

## License

- **live555** - LGPL
- **FFmpeg** - LGPL/GPL
- **OpenCV** - Apache 2.0

---

**If this project helps you, please give it a ⭐ Star!**