idle() the music of ipe()

No. 16· behind-the-music

82 Draft Lines, 15 Cues: What a Clip Keeps

Track 3's lyric draft runs 82 lines per version; its English clip subtitle file holds 15 cues in 58 seconds. We count what the clip files keep and change.

한국어로 읽기 →

How much of a lyric draft actually reaches the screen? Track 3 of dear(you); vol.1, "I Was Scared, Too.", gives us a number to start from. Its English draft has 82 lines. The English subtitle file for its clip has 15 cues, and that clip is a window of 58 seconds. This post lays the two side by side, then runs the same count on the other tracks.

First, the scope. What we read: the lyric drafts for tracks 1 to 10, the 23 subtitle (.srt) files kept beside them, and the creation logs. We do not describe how anything sounds from listening. The subtitle files do not record how they were made, so where they differ from the draft we describe the difference and do not guess the cause. And, as always: the music is produced with AI (Suno). Lyrics are AI-drafted and reviewed by a person.

What the 23 subtitle files are

Two of them belong to track 1 and cover the whole song, one English and one Korean, 25 cues each. Twenty-four cues carry lyrics and the 25th is the closing signature, "ipe. be still", at 3:16 to 3:22. Every time code in these two files is a whole second, and each cue lasts 5 to 8 seconds.

The other 21 files are clips. Their names carry a range such as clip70-128, which we read as seconds into the song, and every one of those ranges is 58 seconds wide. The last cue of each file ends between 55.8 and 58.0 seconds, which fits. Tracks 2 and 3 each have two Korean windows that start a few seconds apart (60 to 118 and 61 to 119; 70 to 128 and 75 to 133). Whether these clips are the ones used for Shorts is not written in any file. The logs list the Shorts links separately.

Track 1: 59 lines become 24 cues

The draft of track 1 has 59 lines in six sections. The English full-length subtitle file regroups them into 24 lyric cues holding 35 text lines. A draft line is a breath; a cue is a phrase. The draft breaks one thought into "Today," / "let me be the first" / "to ask how you've been." and the subtitle shows it as one cue of two lines. The four Korean whispers of the English draft do not appear in the English file at all. So counting lines in one tells you little about the other.

Track 3: a window, not a song

The English clip of track 3 starts with the cue "It was amazing, truly astonishing". In the draft those are two lines, "It was amazing." and "Truly astonishing.", and they are the seventh and eighth lines of a fifteen-line Verse 2. The clip then runs through the Pre-Chorus into the Chorus and ends at 57.78 seconds on "You still learning today" and "Or doing more". The draft reads "You, still learning today, are doing more than enough." So the window cuts a sentence in half, and it holds pieces of three of the seven sections. Verse 1, Bridge, Final Chorus and Outro are outside it.

Inside the window the text is not the draft's text. "Lasts longer than the one who runs the fastest" is spread over three cues, "For the one who walks to the very end", "Last longer" and "And the one who runs the fastest". The Korean whisper "걱정 마…" appears as "Don't worry", in a cue only 0.84 seconds long, the shortest of the 15.

The same lines, cut two ways

The Korean version of track 3 has two windows. The first (70 to 128) has 10 cues and the second (75 to 133) has 12, over largely the same lines. In the first, "괜찮아 그 마음 숨기지 않아도 돼" is one cue. In the second it is two, and "괜찮아" stands alone for 1.4 seconds. So the cue count is a property of the cutting, not of the lyric.

There are also small word differences. The first cue of the first window reads "모두가 눈앞에 펼쳐지더라 신기했어", where the draft has "코드가 눈앞에 펼쳐지더라. 신기했어." The second window starts with "숨기겠어 정말 놀라웠어", where the draft has "신기했어. 정말 놀라웠어." The files do not say whether these are sung variations or transcription errors. Our rule is simple: the draft is the reference, and the subtitle files are a record of a clip.

Counting every clip

Across the 21 clip files the cue counts run from 6 to 18, with a median of 11. The longest is track 4's English clip at 18, and the shortest is track 5's English clip at 6. The studio's length guide for lo-fi suggests keeping about 8 to 12 lines as vocal, and 11 of the 21 clips land in that range. That is as far as we go with it: a cue is not a line, and no document links the guide to the choice of window.

Four creation logs (tracks 2, 3, 4 and 10) also carry an automated note on line counts, for example "SRT 15 lines / lyrics 183 lines (ratio 0.1)". The subtitle numbers match our cue counts: 15 and 12 for track 3, 18 and 9 for track 4, 12 and 8 for track 10, 15 and 8 for track 2. The lyric total we cannot reproduce. Our count for track 3 is 82 per version, 164 for both, and the log does not say how it counted. We read the ratio as an expected flag, since a 58-second window cannot hold a whole song. Track 1's full file runs 3:22.

What we take from it

  • A draft line, a cue and a sung phrase are three different units. Compare them by section, not by count.
  • A clip is a window of about a minute. In track 3 it touches three of the seven sections.
  • The draft stays the reference for wording. The clip files differ from it in places, for reasons the files do not give.

We compared the draft and the clip line by line for track 3 only. The other tracks are counted, not read.

Related