In this lesson
City Pop under a voiceover: leave the speech band open
City Pop keys, guitar and leads sit in the octaves that carry speech. How to arrange and duck a cue so the narration stays clear and the groove keeps moving.
Updated · 7 min read · Lesson 4 of 6 in Genre Guides
In short
- Speech intelligibility lives mostly between 1 and 4 kHz, so keys, guitar and lead lines compete with a voiceover far more than bass and kick do.
- Under narration, dip the music bus in that band, keyed from the voice, instead of pulling the whole cue down.
- Arrange for the edit: keep sung or lead lines for the sections without talking, and bring them back under the b-roll.
In this lesson (8)
The cue that sounds perfect until someone talks
You finish a City Pop cue for a travel vlog: chorused guitar, a bright electric piano, a slap bass line and a lead synth that answers it. On its own it sounds like a late drive through a city at night. The editor drops it under the narration and the first comment on the video asks what the narrator said at 0:42.
The editor's fix is to ride the music fader down until the words come back. By then the cue has lost its bounce, and on a phone speaker the words still smear in places. The level was fine in most of the spectrum. The cue was too busy in the octaves that carry speech.
Why it matters
The City Pop palette sits on top of the words
City Pop, the glossy Japanese pop of the late 1970s and 1980s, is built from sounds with a lot of energy in the midrange: electric piano voicings, clean guitar with chorus, brass stabs, a lead vocal or a lead synth. Those are the same octaves a voiceover needs. The bass line and the kick, which give the genre its pull, sit mostly below them.
So the parts that make a cue sound like City Pop are also the parts that blur the narration. Turning the whole cue down trades away the groove to clear a problem that lives in one region.
The science
Where speech intelligibility lives
The Speech Intelligibility Index (ANSI S3.5-1997) predicts how much of a speech signal a listener can use by weighting each frequency band by how much it contributes to understanding. In its octave-band version, the weights pile up in the middle.
happens band by band: a louder sound raises the threshold for quieter sounds near its own frequency (Fastl and Zwicker, 2007). A slap bass at 100 Hz does little to the consonants of a voice. An electric piano chord at 1 to 3 kHz covers exactly the part of the voice that the index weights most.
Words in the music add a second cost, separate from frequency. Salamé and Baddeley (1989) had people remember lists of digits while music played that they were told to ignore. In their first experiment, both kinds of music hurt recall compared with quiet, and vocal music hurt it more. A second experiment, with more practised participants, found the vocal cost again. Their task was not watching a video, but a viewer following a narrator is also holding words in memory. A sung hook under narration costs more than its level suggests.
That points to two fixes. Take energy out of the speech band only while the voice is talking, and keep sung or busy lead lines for the moments without talking.
Hear the same move on a lead and a pad. The lead's level never changes; only the pad does.
Listen
Free a lead from a crowded pad
A lead and a bright pad share the same range. Cut the pad or duck it under the lead and the lead comes forward without getting louder.
Start with your speakers or headphones low.
Try it
Duck the band, not the cue
You need a City Pop loop, or any busy cue, and 30 seconds of speech: read a paragraph into your phone or use a podcast clip. Step 6 also needs a with an external input. Several DAWs do not ship one, and a keyed from the voice can stand in for it.
0 of 8 steps done
Common mistake
Fixing a band problem with the fader
Pulling the whole cue down until the words come through is the mistake from the hook. The low end and the drums lose their weight first, and the music starts to sound like it is in another room, while the keys can still cover the consonants.
The opposite mistake is a broadband sidechain set too fast. The whole cue jumps up and down with every syllable, and viewers hear the pumping before they hear the music. A slower release, a band-limited dip, or simply a part that sits out are all quieter than the fader.
The third is leaving the topline in. A sung hook or a vocal chop under a talking head competes for attention even when it is mixed low.
Takeaway
Arrange for the edit
When you produce City Pop for someone else's video, write the cue with talking in mind. Keep drums, bass and chords steady, give the lead and the fills the sections without talking, and leave the 1 to 4 kHz region light enough that a dip of a few dB clears it. The cue still sounds like City Pop, and the narrator never has to fight it.
If you are choosing a finished cue rather than producing one, the lesson on streaming versus a creator license covers what you need before it goes under a published video.
Masking
One sound making another harder to hear, usually because they share frequencies at the same moment. A loud sound also masks what follows it for a short time (forward masking). The quieter sound is still there; the ear stops separating it.
Dynamic EQ
An EQ band that cuts or boosts only while the level in its band, or in a key signal, passes a threshold, and stays flat the rest of the time.
Sidechain
Letting one signal control a processor on another, for example the kick triggering a compressor on the bass so the two take turns.
Multiband compression
Splitting a signal into frequency bands with crossover filters and compressing each band on its own.
3 questions
Check yourself
Sources
Show all 3Hide sources
- American National Standards Institute. (1997, reaffirmed 2024). ANSI/ASA S3.5-1997: Methods for Calculation of the Speech Intelligibility Index. Acoustical Society of America.
- Fastl, H., & Zwicker, E. (2007). Psychoacoustics: Facts and Models (3rd ed.). Springer.
- Salamé, P., & Baddeley, A. (1989). Effects of background music on phonological short-term memory. The Quarterly Journal of Experimental Psychology Section A, 41(1), 107-122.
Next lesson · 5 of 6
Cyberpunk Jazz: a human lead on a machine gridThe genre lives on one contrast: synth layers locked to the grid and a jazz lead that leans against it. How to place each layer, and how far to push the lean.Previous: R&B chords that glide: voice leading and a late snare

Written by Virzy Guns
Songwriter and producer, founder of Virzy Guns Production. Now building HealingWave.