Overview

OpenNVR is a self‑hosted video platform that turns any ONVIF or RTSP camera into a full NVR with AI. It runs on hardware you own, a Raspberry Pi 5, a laptop, or a server, and does not depend on cloud services, API keys, or vendor lock‑in. The platform records, streams, and can host any AI model you install, making it suitable for environments where data residency or air‑gap requirements are critical.

To get started, you clone the OpenNVR repository from GitHub, `git clone https://github.com/open-nvr/open‑nvr.git`, move into the directory and launch the system with `./start.sh`. Alternatively, you can pull a pre‑built image that includes the core application. Once running, a web dashboard appears where you add cameras by entering their ONVIF or RTSP URLs. You can also use a smartphone, a laptop webcam, or a local video file as a camera source if no physical cameras are available. After a camera is online, you can view a live feed, set recording parameters, or draw zones for specific object classes. The interface lists each camera's status, health, and last event timestamps.

OpenNVR's AI capability is plug‑in based. Any model that follows a simple open contract can be dropped into the system. The model's weight file is fingerprinted and stored, and every inference event is logged with that fingerprint, making it possible to trace alerts back to the exact model version. Ready‑made apps, such as person detection, vehicle tracking, or under‑occupancy alerts, are packaged as containers that can be installed with a single click. Developers can also build custom apps or extend the provided SDKs. When an event occurs, say a person track or a vehicle passage, the system pushes the detection as a message to an event bus that you own. You can trigger downstream actions, like sending an email or logging the event, through that bus.

The platform maintains health monitoring for every layer: camera connectivity, recording, AI inference, and the event bus. If any component fails, the dashboard displays an alarm and a description of the issue, allowing quick troubleshooting. All recorded footage, AI outputs, and inference logs stay on the local device, permitting air‑gapped deployment. The system also re‑streams cameras that are isolated on a private network over TLS, ensuring secure playback on the host machine.