Four episodes, two of them unseen

Widened to ten seconds either side, as you asked. Episodes 5 and 6 are produced. Episodes 1 and 2 are the real test — nobody has ever listened to them, and their spans came from the detector alone.

Why the detector was allowed to be generous. A false positive used to mean processing audio that did not need it. Here it means replacing clean room with other clean room from the same recording, which costs almost nothing. So the threshold over-flags rather than under-flags — the opposite of every earlier round, and the right trade for this method. If a region comes back nothing was wrong here, that is a cheap miss rather than damage.

Opener offsets, measured not assumed: episode 1 opens with the voiced 10.62 s opener, the other three with the 7.03 s sting. An earlier version of this page assumed 7.03 for all four, so every episode-1 comparison was 3.59 s out of step and the two buttons played different moments. The offsets are now read from the files.

The voice stem is untouched in all four — verified bit-identical — so nothing about widening or over-flagging can reach the voices.

nothing playing

Start with episodes 1 and 2 — those are the ones that tell us whether this works without you in the loop. Each clip plays the whole widened region.

Episode 1

spans found automatically — nobody has listened to this episode. 17 spans became 8 regions after widening, covering 194 s of 1046.

1 0:07.03 13 s region

2 1:42.27 21 s region

3 4:26.10 27 s region

4 11:17.70 23 s region

5 11:43.51 30 s region

6 12:13.94 29 s region

7 15:23.79 29 s region

8 17:09.54 23 s region

Episode 2

spans found automatically — nobody has listened to this episode. 13 spans became 4 regions after widening, covering 97 s of 750.

1 0:07.03 20 s region

2 6:03.57 28 s region

3 8:59.84 23 s region

4 12:10.90 26 s region

Episode 5

spans you confirmed. 8 spans became 5 regions after widening, covering 139 s of 432.

1 0:07.03 23 s region

2 1:38.09 27 s region

3 2:18.34 41 s region

4 4:19.22 22 s region

5 6:45.51 25 s region

Episode 6

spans you confirmed. 3 spans became 3 regions after widening, covering 68 s of 550.

1 0:13.22 23 s region

2 1:39.53 25 s region

3 8:57.01 20 s region

Your answers appear here — copy and send back.