Clairva annotates video at frame-level fidelity — objects, hands, depth, scene, speech, behavior — across multiple model families and content domains.
