real-time video intelligence · self-hosted

vidarax_

Vidarax is a self-hosted video intelligence engine. Give it a live stream or a recording and it returns one ordered stream of markers, descriptions, and structured events. Your application works with data instead of owning a media stack.

every frame is checked. selected moments become durable, queryable events.

How it works

01

ingest

Upload an MP4 or attach WebRTC, RTSP, or HLS. The source adapters differ, but every accepted frame enters the same bounded processing graph.

02

decide what matters

A deterministic gate checks scene cuts, flicker, ghosting, exposure shifts, and loops before inference. Optional novelty signals can reduce repeated analysis, but the engine reports its decisions instead of promising one universal savings number across unlike streams.

03

analyze only that

Selected frames reach the configured vision provider. Tiered inference can ask a smaller model first and escalate when confidence is low. Each result joins the ordered event log and is available through REST, streaming APIs, and the typed SDK.

mp4 webrtc rtsp / hls ingest one pipeline decide every frame quiet, set aside what matters analyze small model first, larger if unsure events rest / streaming

Selectivity and end-to-end throughput depend on the stream, provider, and hardware. Vidarax exposes the decisions and timings needed to measure that deployment directly.

Built for operators

where it is aimed: gameplay QA surveillance review manufacturing inspection robotics logs

Early access

Vidarax is in development. The documentation covers the API, ingest paths, deployment boundaries, and the internal media plane. For the full argument, read A video stream through Vidarax. If you want to test it against your streams, write to cosmin@bararu.com and say what the streams are and what events you need out of them.