Shmili episode 2 — what should come out

Episode 2 — the noise pass. Tap the sound on its own first; if you cannot tell what it is, tap in context. Pick a button or write your own words. Then hit copy at the bottom and send me the block.

Before the ruffles — something that needs your ear more than they do.

Episode 2's working file had 10.0 seconds of your speech destroyed, in 28 places. Not noise on top of it — the speech itself replaced by digital silence. The worst is 2.4 s at 11:04.40, where the recording has you at −10 dB and the working file has nothing at all. It is in the episode that is live right now.

It was the old denoiser, the same one that ate your welcome line in episode 44. Episode 2 was never checked for it, because the check compares against the original and episode 2 is the one episode with a hand trim, which threw the comparison off. The trim turned out to be a clean 14.100 s off the front, so the comparison works after all.

I put your real recording back in all 28 places. Nothing was regenerated — it is your voice from the original tape. Have a listen to the worst one:

What I need from you here: does the restored speech sound right, and is the room tone behind it too loud against the rest? The restored bits carry the original room sound, which is noisier than the cleaned-up audio around them. If it bothers you, say so and I will match it.

About the 32 sounds below.

No volume floor — you retired that. I scanned to the bottom of the room channel (quietest block -97 dB, median -69 dB) and found 123 separated candidates. The 32 here are the loudest, from -15.3 dB down to -55.9 dB.

91 more were left off, between -55.9 and -86.1 dB. That is a limit on your time, not a judgement that they do not matter — on episode 1, 21 of 22 candidates that quiet turned out to be real. Say the word and I will put them in front of you too.

Each sound now carries a measured length, not just its loudest instant — 7 of these run longer than a second. Where the sound is too long for the three-second clip there is an extra button for all of it.

Not on this list, on purpose: the 28 repaired places above. They are the loudest things in the room channel now, and 5 of the top 6 would have been on this page — but they are your restored voice, not ruffles. Marking one "cloth" would delete it again.

nothing playing

Episode 2 32 sounds, loudest first

T02 1 7:29.00 1.0 s long · -15.3 dB

T02 2 8:27.50 0.5 s long · -45.1 dB

T02 3 8:33.00 0.2 s long · -47.7 dB

T02 4 3:04.75 0.5 s long · -50.6 dB

T02 5 4:20.00 1.0 s long · -50.7 dB

T02 6 4:24.25 2.5 s long · -51.1 dB

T02 7 3:26.50 0.8 s long · -51.6 dB

T02 8 1:40.50 0.5 s long · -51.9 dB

T02 9 0:02.00 0.5 s long · -52.0 dB

T02 10 0:08.25 0.5 s long · -52.5 dB

T02 11 0:33.50 0.8 s long · -52.8 dB

T02 12 0:16.75 0.2 s long · -52.8 dB

T02 13 5:48.25 2.0 s long · -52.9 dB

T02 14 7:57.00 1.2 s long · -53.2 dB

T02 15 12:12.25 0.5 s long · -53.3 dB

T02 16 9:41.00 1.2 s long · -54.0 dB

T02 17 11:57.00 0.5 s long · -54.1 dB

T02 18 1:35.25 0.5 s long · -54.2 dB

T02 19 4:12.00 0.8 s long · -54.2 dB

T02 20 0:21.00 0.2 s long · -54.5 dB

T02 21 9:33.50 0.8 s long · -54.5 dB

T02 22 6:22.50 0.2 s long · -54.7 dB

T02 23 4:35.00 1.0 s long · -54.9 dB

T02 24 3:31.25 0.8 s long · -55.0 dB

T02 25 1:00.00 1.5 s long · -55.2 dB

T02 26 4:46.75 1.2 s long · -55.4 dB

T02 27 5:21.50 2.0 s long · -55.4 dB

T02 28 1:17.25 0.2 s long · -55.4 dB

T02 29 1:29.25 0.5 s long · -55.5 dB

T02 30 5:13.25 0.8 s long · -55.6 dB

T02 31 4:57.00 0.5 s long · -55.9 dB

T02 32 1:11.25 0.5 s long · -55.9 dB

0 of 32 answered

Your answers appear here — copy the whole block and send it back.