English by a Woman, Korean by a Man — The Rule That Splits the Two Versions, and How the English Was Paraphrased

한국어로 읽기 →

All ten songs on dear(you); vol.1 are made in two versions, and the album plan separates them in a single line: a sweet-voiced female for the English-main version, a gentle male for the Korean-main version. The rule that puts the opposite language in as whispers was read across all ten songs in Whispers in the Other Language. This post reads the side that one left alone: why the main language and the vocal gender are only ever written down as a pair, and what the principle of a "natural paraphrase" (our translation of the plan's wording) for the English looks like inside the lyric lines. The order in which a song is made is covered in How a Song Gets Made, so it is not repeated here.

First, the scope. What this post read is the album plan, the song design documents (the lyric drafts and the Suno style field), the music production packages and the production logs. Everything said about the vocals is a specification written in those documents, not an impression from listening to the recordings. We did not check whether the recordings sound the way the specifications say. And, as always, the music is produced with AI (Suno). Lyrics are AI-drafted and reviewed by a person.

A pair written in one line

The pairing is written in two places in the album plan and once more on the song status board. In the overview table, the version field reads "two versions per song" (our translation) with "English Main" and "Korean Main" side by side, followed by a Korean description of the voice: "감미로운 여성" (a sweet-voiced female) and "다정한 남성" (a gentle male). In the production-rules table, the English-version row holds "English (natural paraphrase, no literal translation)" (our translation) and the female voice in one row, and the Korean-version row holds "Korean (the original as delivered)" (our translation) and the male voice. The status board carries a similar table under the title "dual-version rule (confirmed 2026-07-04)" (our translation).

The song design documents carry the same pairing. The version/vocal field in the header table of each of the ten is character-for-character identical (we checked all ten with a script), and the two headings that hold the lyrics read "English Main (sweet female vocal)" and "Korean Main (gentle male vocal)" in every one. The hand-off memo that sends each song to composition also names the same pair in all ten.

What stands out is that the pair never appears as two separate choices. No document has one field for the main language and another for the vocal gender, and a male main vocal on an English-main version, or a female main vocal on a Korean-main one, is written nowhere in the ten songs.

The closest thing to a reason

So where is the reason for this pairing? The album plan and the song design documents give the rule and the voice names, nothing more. The closest thing to a reason is two lines in the selection-criteria field of the music production log for song 01 (our translation):

EN version: a sweet female vocal to achieve a warm sense of comfort

KR version: a gentle male vocal to achieve a friendliness that fits Korean sentiment

These lines say what each voice was expected to bring. They do not say why the English side got a woman rather than a man, or whether there was a reason not to swap the two. Only song 01 has such a field. The logs of the other nine have no selection-criteria field, and in the 01 log the other fields (the style string entered into Suno, and the lyric revisions) are left unfilled.

The specification is fixed; the wording drifts

The vocal specifications of the ten songs, copied from the song design documents.

What stays fixed across the ten songs is the pairing itself. The wording moves in three places.

• On the female side, nine songs say "sweet warm female vocal"; only song 05 says "breathy sweet female vocal."
• On the male side, songs 01 to 04 say "gentle warm male vocal." From 05 on, "whisper" is added: "whisper gentle male vocal" for 05, 07, 09 and 10, and "warm whisper male vocal" for 06 and 08.
• An opposite-gender harmony is specified for eight songs: on the Chorus for 03 and 04, on the Final Chorus for 06 to 10, with no position given for 05. The specifications for 01 and 02 do not mention it.

One thing follows from the third point. As specified, the English-main (female) version also carries a male harmony, and the Korean-main (male) version a female one. "English is female" is a statement about the main vocal, not about the whole version.

The music production packages hold the English phrases written for pasting into Suno's style field. For the female side, all ten contain "warm husky female vocal, slightly smoky tone, mid-low range"; for the male side, all ten contain "deep husky male vocal, warm baritone, gentle Korean male voice." Only the male side gains "whispered delivery," from song 05 on. How the Korean labels "sweet" and "gentle" relate to husky, smoky and baritone is not written anywhere. The male phrase also names a language ("Korean male voice"), while the female phrase never says English in any of the ten songs. What that asymmetry means is not on record either. We could not confirm that the package phrases were entered as written: the log field for the entered string is empty for song 01, and the other logs have no such field. The tempo is the same for both versions: the song design document has a single tempo field, and the BPM in the packages is the same for both.

The original sung by a man, the paraphrase by a woman

Overlay the second rule on a pairing that only ever comes as a pair, and one result follows. The Korean side is the original and the English side is the paraphrase, so the original is always sung by the male voice and the paraphrase always by the female voice.

The song memos say the same. The memos for songs 01 and 02 give the starting point as "only Korean was handed over" and go on to "English is generated as a natural paraphrase" (our translation), and the composition-request memos for songs 02 to 10 all read "Korean original + English paraphrase" (our translation). The Korean is the text given first; the English is a text built on it. Being the original, though, does not mean untouched. The album plan's checklist has, next to "English translation naturalness check done (smoothing awkward phrasing + easing repetition devices)," the line "Korean original check done (0 typos, chorus hook tidied, bridge refrain unified)" (our translation, both).

Whether this overlap was intended is not on record. What the drafts confirm is that it runs the same way in all ten songs.

What is written about paraphrase

Three things are written about paraphrase.

• The album plan's production-rules table: "English (natural paraphrase, no literal translation)" (our translation).
• The reason, in the multilingual operating-principles document (last updated 2026-06-26): "Don't sing an AI literal-translation lyric as it is → rhythm and feeling break, and brand quality suffers." (our translation)
• The album plan's checklist: "English translation naturalness check done (smoothing awkward phrasing + easing repetition devices)" (our translation).

There is nothing beyond that. Which lines were smoothed, which repetition "easing repetition devices" refers to, and whether a criteria sheet was kept are not on record. The lyric-revision field is unfilled in the 01 log, and in the logs for 02 to 10 it holds only an auto-generated instruction line. So from here on, this is not quotation from documents but a reading of the two lyric drafts placed side by side line by line. It is a tidy-up by the reader, not a recorded rule.

The skeleton stays

The easiest thing to count is the skeleton. Across the ten songs, each version has 68 lyric sections, and in every song the English and Korean section names and their order are identical. In 55 of those 68 sections the line count is also the same; the other 13 differ by exactly one line, and in ten of those thirteen the Korean has the extra line. (Titles and blank lines are excluded; whispers and the closing signature are counted.) Even as a paraphrase, inside the section frame the line count is either the same or differs by a single line (we did not count whether the line breaks fall in the same places).

What happens inside a line

Every line is taken from the lyric drafts, with the draft's line breaks joined by spaces (one line in the picture can be several lines in the draft). The numbers on each card are the songs the lines come from. The grouping is ours, from reading them.

Inside the lines, five shapes appear.

One: the work done by the sentence ending is carried over. The title of song 01, 「오늘도 많이 힘들었지?」, is "Hard Day Again, Wasn't It?" in English. If the Korean ending 「~지?」 asks for agreement, "Wasn't it?" does the same job. The Korean ending 「~잖아」 (roughly "as you know") closes twelve lines in the Korean lyrics, and three of them become tag questions. 「사실은 조금 속상했잖아.」 becomes "but the truth is, it hurt a little, didn't it?"; 「그 말이 입에 먼저 나왔잖아.」 becomes "Those words came out first, didn't they."; and 「한 번쯤은 했잖아.」 becomes "you've had it once, haven't you?" The other nine are plain statements. For example, 「한 번쯤은 하게 되잖아.」 is "everyone has it once in a while." Promises and invitations work similarly. In song 01, 「오늘은 내가 먼저 네 안부를 물어볼게.」 becomes "Today, let me be the first to ask how you've been."; and in song 05, 「노트북을 천천히 닫아볼까. 휴대폰도 잠시 뒤집어 둘까.」 becomes "Shall we slowly close the laptop? And turn the phone face down for a little while?" The shape of the ending changes, but the job of the line (asking for agreement, promising, inviting) stays.

Two: the direction of a question turns over. In song 09, Verse 2, the Korean is 「조금 늦게 걸어도 괜찮지 않을까. 조금 천천히 웃어도 괜찮지 않을까.」 (roughly, "Wouldn't it be okay to walk a little late? Wouldn't it be okay to smile a little slowly?"), and the English is "would it be so bad to walk a little late? Would it be so bad to smile a little slowly?" The Korean asks whether it is okay; the English asks whether it is bad. The direction is reversed, but the twice-repeated structure and the work of reassuring are the same.

Three: a word changes with its place. The word 「고생」 (roughly "hardship endured") appears in six lines across four songs. In song 04, 「오늘도 고생했어요.」 becomes "You worked hard again today."; in song 06, 「오늘도 고생 많으셨어요.」 also becomes "you worked hard again today."; in song 08, 「고생했다고 말해줘」 becomes "you did well." Song 10 uses the same line, 「정말 고생했어.」, three times, and in English it is "did so well today," which is also the title, "You Did So Well Today." One Korean word shows up as three English sentences. It reads less as swapping a word for a word and more as rewriting, in English, what is being said at that spot.

Four: a layer only Korean has is unpacked or dropped. In song 05, 「쉼표가 있어야 노래가 아름답듯」 becomes "Just as a song is beautiful only because of its rests." The Korean 쉼표, which can mean a comma or a musical rest, becomes the musical term "rests." Also in song 05, 「기다릴 줄 아는 아이들이야」 (literally "they're kids who know how to wait"), said of the unread mail and the unfinished code, becomes "they're the kind that know how to wait a little." The phrase 「내일의 우리가」 ("tomorrow's us") appears three times, in the Outro of 05, the Final Chorus of 08 and Verse 2 of 10, and all three English places read "Because tomorrow, we ...": the "tomorrow's" layer shrinks to the word "tomorrow." The English also adds things. In song 05, 「오늘 하루는 여기까지.」 becomes "This day of ours — this far is enough."

Five: the same sentence is kept the same. The album plan says the Bridge sentence shared by songs 04 and 07 was unified in both English and Korean. The Korean is 「AI는 좋은 답을 만들 수 있어. 하지만 좋은 하루를 만드는 건 언제나 사람이더라.」 and the English is "AI can make good answers. But the one who makes a good day was always a person." Both songs carry it, identical in each language. The Korean ending 「~더라」, which reports something learned by experience, seems to have its counterpart in the past tense of "was always." That is the reader's interpretation. The two songs are read side by side in The Same Sentence in Two Songs.

Same song, different length

The lengths of the two versions of each song. These are file lengths, not listening impressions.

The lyric skeletons are nearly the same, but the recordings are not the same length. The album compile plan (a draft) lists measured lengths for the ten songs (taken with a tool that reads the length of audio files): 33 minutes 49 seconds in total for the English-main versions and 39 minutes 47 seconds for the Korean-main ones. The Korean-main version is longer in nine songs, and only song 05 is a second longer in English. The biggest gap is song 10, at 1 minute 17 seconds (3:21 against 4:38), and in total the Korean side is about six minutes longer. For songs 02 to 10 we re-measured the audio files and they matched the document to the second after rounding. Song 01 has only a video file, not an audio file, so we could not confirm it the same way.

Why they differ by this much is not on record. We note only that the tempo is the same for both versions and the line counts are nearly the same.

What is not on record

• Why the main language and the vocal gender were tied into one pair, and why the pairing was never swapped. The two lines in the 01 log say what was expected, nothing more.
• How the Korean labels "sweet" and "gentle" relate to the husky and baritone phrases in the packages; whether the package phrases were actually entered into Suno; why only the Korean-side phrase names a language.
• Whether a criteria sheet was kept for the paraphrase, which lines were revised, and which repetition "easing repetition devices" points to.
• Why the two versions differ in length.
• Whether the female and male vocals sound as specified, how the opposite-gender harmony sounds, and whether the sung portion matches the draft.

This post does not write any of these as if they were recorded. What we confirmed is that the same pair is written across all ten songs, where the specification wording moves, and how the two drafts line up, skeleton and line by line.

Listen

The English (female) and Korean (male) versions of a song can be played back to back to follow how the lines pair up. All ten songs are on YouTube; below are the ones quoted most in this post.

• Song 02: It's Okay, Take Your Time. (괜찮아, 천천히 말해봐.) — English version · Korean version
• Song 05: It's Okay to Rest a While. (잠깐 쉬어가도 돼.) — English version · Korean version
• Song 09: Let's Smile a Little Tomorrow. (내일은 조금 웃자.) — English version · Korean version
• Song 10: You Did So Well Today. (오늘도... 정말 고생했어.) — English version · Korean version

Once more, plainly: the music is produced with AI (Suno). Lyrics are AI-drafted and reviewed by a person. You can also listen on the idle() home page.