Identity captured at source, not inferred from pixels. A deterministic metadata layer for video production that binds physical object identity to pixel-space representation at the moment of filming.
Attach RF tags (UWB preferred) to physical objects on set. Each tag carries a unique identifier and provides real-time spatial position via ranging.
Fixed anchors around the set define a world coordinate frame. The localization engine computes 3D positions for each tag at 10–100 Hz with centimeter-level accuracy.
The RFML Bridge fuses RF positions with camera pose, lens calibration, and SMPTE timecode to project each object into pixel space — frame by frame, in real time.
Per-frame metadata records carry object identity, world-space position, and pixel-space bounding regions. Sidecar files travel with the video from set to screen.
{
"timecode": "01:02:03:12",
"frame": 110592,
"fps": 24.0,
"camera": {
"serial": "CAM_01",
"pose_world": { "x": 1.234, "y": 0.456, "z": 1.789 },
"focal_mm": 35.0
},
"objects": [{
"object_id": "JKT_0142",
"tag_id": "uwb:0xA17C9B2E",
"world_pose": { "x": 3.12, "y": 1.07, "z": 0.92 },
"pixel_bbox": { "x": 812, "y": 214, "w": 188, "h": 402 },
"confidence": 0.91,
"sku_refs": ["SKU_881223"]
}]
} // one record per frame — world-space + pixel-space per object
Viewers select objects to purchase. Identity is exact — SKU-level, not category-level.
Automated tracking of which objects appear in which shots across an entire production.
Object-level engagement metrics, conversion funnels, and temporal trends.
Per-object licensing constraints evaluated at playback — by territory, time window, or channel.
Deterministic player and gear identification for real-time overlays and statistics.
Object identity and pose streams enable AR overlays on companion devices.
RFML is in active development. If you work in production technology, interactive video, or shoppable media, I'd like to hear from you.