Speaker: ×Tip: click a speaker to filter⚠ Operator access temporarily unavailableThis can happen if the PIN was changed, or if your network has had too many failed sign-in attempts recently. Retrying every second; access is restored automatically once available. If the PIN was changed, re-enter it above.AdminMonitorDetector model
Filter music/singing
Auto-stop GPU
Save clipsSave non-target clips
VAD/AED gate monitor
VAD / AED gate monitor
Waiting for a live call…
VAD threshold: —Polling every 1s · newest first
Open via the operator link (?token=) and enable the toggle; rows populate during a live call.
Time (SGT)
Speaker
VAD
Speech
Singing
Music
Silence
Sent
Enable the toggle above to start monitoring.
Meeting
Monitoring
—
Detector
—
Starting takes ~2 min. Scoring begins once Ready.
⚠ Detector running with no active call · billing until Stop
No active call. The detector will start billing now and keep billing until you press Stop.
Call
—
Active speaker
—
Detector uptime
—
Latest result
Waiting for audio…
Authentic
—
Suspicious
—
Speaker judgement
Speaker
Assessment
Clips
Spoof%
No speakers yet.
Detection log
Time (SGT)
Speaker
Result
Confidence
No detections yet.
Session labeling & model comparison
0 / 0 labeled
Select a clip to play
No session loaded.
keys: b/s/u label current clip + advance · ←/→ or ↑/↓ move · space play/pause. Rows run in meeting-audio order and # is that position; the #n beside the speaker is the bot's per-speaker clip index, which is what the labels and the CSV are keyed by. Each model column shows that detector's verdict for the clip; a blank cell means the model was not run. The Speaker filter only changes what is shown: labels and the export always cover the whole session. Labels save to the bot as you go; Export writes a CSV on the VM.