Episode 6 uses the spans you drew by hand. Episode 5 uses the ones you
confirmed from an automatic search — so it is also a test of whether the boundaries hold
up when you are not the one drawing them.
What is the same in both. One gain across the whole spectrum, lowered where there
is no voice and back to normal where there is. No narrow bands cut, so the metallic
ringing cannot come back. Gain never rises, length is unchanged, and everything outside
the marked spans is bit-for-bit identical to the original.
Where they differ, and it matters. On episode 6 you drew the boundaries and every
span got a real reduction. On episode 5 the boundaries are mine, and three of the eight
came back with no reduction at all — the audio there was already at room level, which
usually means the noise is just outside the span I was given. Those three are marked in
amber. If you hear the rustle in them, that is the boundary problem again rather than the
processing.
nothing playing
Episode 6 — the three spans you marked
You drew these yourself on the marking page. Each starts 1.5 seconds early so you hear the run-in.
10:23.223.2 s · noise down 16 dB · 21% held as speech
21:49.534.6 s · noise down 8 dB · 15% held as speech
39:07.013.1 s · noise down 8 dB · 0% held as speech
Episode 5 — the eight you confirmed
These come from the ten candidates you marked as cloth rustle, with touching ones merged. They were found automatically, so unlike episode 6 the boundaries are still my guess — three of them turned out to be already at room level and got no reduction at all.
10:15.314.9 s · noise down 20 dB · 0% held as speech
21:48.092.5 s · noise down 17 dB · 0% held as speech
31:53.072.5 s · no reduction — already at room level · 100% held as speech
42:28.342.5 s · noise down 20 dB · 20% held as speech
52:32.902.5 s · noise down 20 dB · 0% held as speech
62:47.292.5 s · noise down 20 dB · 11% held as speech
74:29.222.5 s · no reduction — already at room level · 50% held as speech
86:55.514.5 s · no reduction — already at room level · 14% held as speech