Event Raptor
Domain-specific Vision-Language Model for real-time security, safety, and smart infrastructure intelligence
Book a demoRelated materials
FAQ
Event Raptor is Scylla AI's proprietary Vision-Language Model (VLM) developed entirely in-house and built on ScyllaNet, Scylla's foundational computer vision architecture. It has been designed specifically for physical security and surveillance environments. Unlike general-purpose AI models, Event Raptor was trained on Scylla's proprietary real-world surveillance datasets spanning diverse environments, camera types, lighting conditions, and threat scenarios, making it purpose-built for operational security rather than adapted from consumer AI.
See also: Event Raptor
Event Raptor is a domain-specific VLM trained exclusively on real-world security surveillance data, which means it understands the specific visual language of security environments: occlusions, low light, crowded scenes, and high-stakes threat scenarios where hallucinations are unacceptable. It runs entirely on-premises and in air-gapped environments with no reliance on third-party APIs or external cloud infrastructure, giving organizations full control over their video data and security intelligence.
Event Raptor recognizes and describes more than 1,000 object, behavior, and situational classes, enabling detailed scene understanding far beyond basic object detection. The model generates operationally relevant descriptions of security events, providing security teams with immediate context and actionable intelligence rather than raw detection outputs.
Event Raptor supports real-time inference with sub-100 millisecond latency per frame, enabling live security analysis without perceptible delay. This performance is achieved through ScyllaNet's architecture, which is optimized using genetic algorithms to deliver high-performance visual understanding while maintaining a lightweight footprint suitable for edge deployment.
See also: ScyllaNet
Yes. Event Raptor was specifically designed for deployment in air-gapped environments where external data transmission is restricted or prohibited, with no reliance on third-party APIs, external cloud infrastructure, or internet connectivity. This makes it particularly suited for government, defense, critical infrastructure, healthcare, and financial institution deployments where data sovereignty and network isolation are mandatory requirements.
Event Raptor is integrated into Scylla AI's core platform and Alarm Hub, where it supports forensic video search via natural language queries, real-time alarm enrichment with contextual descriptions, and the creation of custom threat recognition rules using plain language; all without requiring coding or cloud connectivity. Through Forensics Pro, Event Raptor enables investigators to search archived and live footage using descriptive text prompts or reference images, returning ranked timestamped results across the full camera network.
Event Raptor is optimized for the conditions that challenge standard AI models most: low-light environments, crowded scenes, partial occlusions, dynamic backgrounds, and diverse camera types and angles; the real-world conditions consistently encountered in operational security environments. Its training on proprietary security surveillance datasets rather than general internet imagery, which means it performs reliably in the scenarios security teams actually face, rather than curated benchmark conditions.
Event Raptor automatically enriches real-time alarms with contextual natural-language descriptions telling operators not just that something was detected, but what happened, who or what was involved, and why it is relevant, dramatically reducing the cognitive load of alarm triage. By enabling plain-language forensic search and custom threat rule creation without coding, it also eliminates the need for specialized technical expertise to configure and operate advanced AI video analytics.