1. Upload a clip
Drop a video here or click to choose
mp4 / mov / mkv / webm — up to 150 MB
2. Options
Cloud VLM: AXWEL AI tier-1/2 (nuanced, per-clip cost).
Local: on-box signal fusion (live face recognition, zone intersection, audio full spectrum, transcript keywords, dwell, time context) — free and deterministic.
Both: runs both in parallel, returns side-by-side with an agreement report.
Adds external + insider AXWEL AI vision calls + local Whisper STT when available.
Only for diagnostics — won't affect production gating.
Keeps the legacy side-by-side with raw person-detect + transcription, independent of the structured local analyzer.
Local vs Cloud VLM — 30-day agreement
No comparison data yet. Run Both mode to accumulate history.
Local signals
loading…
Flip a signal off when the underlying data isn't populated (no zones, no pose backend, etc.). Applies immediately — no restart.
Weight calibration
No calibration runs yet. Requires human-reviewed incidents.
Replay past decisions
Current config
- Loading…