Measure audible onset at the first played speech sample

When response latency is defined by audible speech, a packet containing speech is insufficient evidence for its onset. Locate the first sample above the documented speech-energy threshold and add its within-packet offset to the playback clock. Exclude any tail discarded before that sample plays. Keep transport receipt, synthesis first byte and audible playback as distinct landmarks. An amplitude threshold remains a heuristic: fades, breaths and noise can move the boundary, so preserve the threshold, waveform hash and raw landmarks rather than calling it human-perceived ground truth.

Applies when: Audio benchmarks score response time from packetized or synthesized output. Keywords: audible onset speech latency packet silence sample threshold playback