Frigate NVR

Analyzes video streams in real-time using local YOLO models to identify and classify objects such as people, vehicles, animals, and packages. Processing occurs on local hardware with sub-second latency, eliminating cloud dependencies.

Accepts and manages video feeds from multiple RTSP/ONVIF compatible cameras, handling stream normalization, buffering, and coordination across unlimited camera sources. Supports diverse camera hardware ecosystems.

Enables searching through historical detection events by object type, timestamp, camera, and confidence level. Allows users to quickly find specific incidents or patterns in their security footage.

Supports standard RTSP and ONVIF protocols for camera communication, enabling compatibility with a wide range of IP cameras from different manufacturers without proprietary integrations.

Provides complete access to source code, allowing technical users to customize detection models, add features, and modify behavior to suit specific needs. Enables community contributions and transparency.

Operates entirely without recurring subscription fees or cloud service dependencies. All functionality is available through one-time setup on user-owned hardware with no ongoing costs.

Stores video recordings locally on user-controlled hardware with configurable retention policies. Eliminates cloud storage dependencies and provides unlimited storage capacity based on available disk space.

Captures and stores video clips and snapshots when specific detection events occur, such as person detection or vehicle arrival. Reduces storage overhead by recording only relevant moments rather than continuous footage.

Sends real-time alerts and notifications when detection events occur, with support for multiple notification channels and customizable alert rules. Integrates with Home Assistant and other automation platforms.

Provides a web interface for viewing live camera feeds and reviewing recorded video footage. Allows users to scrub through recordings, search by timestamp, and access footage from any device with a browser.

Integrates deeply with Home Assistant to expose camera feeds, detection events, and controls as entities. Enables creation of automations and scenes based on detection events within the Home Assistant ecosystem.

Processes all video analysis and storage entirely on local hardware with no data transmission to external servers or cloud services. Ensures complete data privacy and ownership without third-party access or data selling.

Leverages GPU hardware acceleration to perform real-time object detection inference at high speed. Enables sub-second processing latency for multiple concurrent video streams without CPU bottlenecks.

Allows users to define specific regions of interest within camera feeds where detection should occur, and mask out areas to ignore. Reduces false positives and focuses detection on relevant areas.