Frigate

Frigate is an open source network video recorder (NVR) for IP cameras built around real-time, local AI object detection. It processes camera feeds on your own hardware to reduce false positives and enable fast, searchable event review without relying on cloud analysis.

Key Features

• Real-time object detection using local accelerators (GPU/TPU) to distinguish people, cars, and other objects from motion
• Efficient motion detection to decide when and where to run inference, minimizing resource usage
• Continuous (24/7) recording and event recording with retention policies based on detected objects
• RTSP restreaming to reduce the number of direct connections to cameras
• Low-latency live viewing with WebRTC and MSE support
• Zone-based filtering and object tracking to refine events and alerts
• MQTT integration for automations and interoperability, plus tight Home Assistant integration via a custom component

Use Cases

• Build a privacy-preserving home security camera system with local AI detections and recording
• Trigger smart-home automations (lights, alarms, notifications) based on specific detected objects via MQTT/Home Assistant
• Reduce time spent reviewing footage by focusing on object-based events instead of raw motion

Limitations and Considerations

• Best results typically require a supported AI accelerator (GPU/TPU); CPU-only setups may have limited throughput
• Requires careful camera and zone/mask configuration to balance performance, accuracy, and storage usage

Frigate combines high-performance local detection with recording and review workflows to create a customizable, locally controlled security camera platform. It is especially well-suited for users who want low-latency viewing and automation-friendly integrations while keeping video processing on-premises.