Live Fact Check
AudioJevBench showcase

AudioJevBench · live audio demo

Three seconds in.
One label out.

Speak naturally. A decision model listens to the rolling audio window and answers A, B, or C — about five times a second.

Live verdict
READY
Tap the mic or play a spoken sample. Each check hears a rolling 3-second window.
Recent label agreement share of recent A / B / C calls
A · True – B · False – C · Not a claim –
Use your microphone
or — no mic needed
60s
The rolling verdictone mark per 200 ms tick · color = label · height = recent agreement
truefalsenot a claimin flightsilence (skipped)
Model latency p50
–ms
Last call
–ms
Requests / s
0.0
In flight
0/ 12
Samples — played through your speakers and fed into the exact same pipeline