Scoring under voiceover is not a matter of turning the music down. If narration and score carry the same rhythmic detail, attack at the same moments and occupy the same register, a lower fader only makes the collision quieter.
I treat the voice as the scene’s lead line. The score can shape its direction, expose a contradiction or connect images the words cannot join. It should not force the audience to choose between following the sentence and hearing the musical idea.
That balance starts in the brief and the arrangement. The dub is the final opportunity to refine it, not the first time anyone should notice the problem.
Define what the narration is doing
Voiceover can deliver facts, reveal private thought, compress time, misdirect the audience or create distance from the image. Those jobs need different musical responses.
I ask production which lines carry information that cannot be recovered elsewhere. I also mark changes in point of view and the emotional destination of the sequence. If every sentence is treated as equally important, the music has no useful hierarchy to follow.
A strong brief identifies what the viewer must understand, what the image already supplies and what remains emotionally unresolved. A temp track may demonstrate energy or texture. It cannot answer those editorial questions.
Compose around speech rhythm
Speech already contains rhythm. Consonants create attacks. Breaths divide phrases. Pauses open space. A light arrangement with constant transients can fight narration more than a sustained, darker texture.
I look for pockets rather than covering the whole cue with activity. A figure can answer the end of a sentence, expand through a breath or simplify as the language becomes denser. Silence is part of the arrangement.
Dry percussion, fast ostinatos and exposed melodic lines are not automatically wrong. They need a defined lane. If they run continuously beneath detailed narration, the mixer may have to remove the very material that gave the cue its identity.
Writing with those priorities early preserves character. Trying to rescue an overcrowded cue with level automation usually produces a less convincing version of the same conflict.
Label scratch VO as scratch VO
Editorial often builds the cut with a scratch read. It provides timing and lets the team test structure. It does not lock the pace, breath pattern or performance of the final narration.
I want the file and picture version to state that status clearly. I also want to know when the final record is scheduled. If I am writing before that date, I avoid making every musical hit depend on a temporary pause.
A common failure is beautifully precise: the cue rises inside a breath in the scratch read, then the final narrator connects the two phrases. The new performance has not damaged the music. The cue was tied too closely to material everyone knew would change.
When final VO arrives, I review the full sequence. I do not mechanically chase every new breath. Some independence between score and voice creates flow. I revise the moments where a hit obscures a word, changes emphasis or makes the narration feel rushed.
Spot priorities, not just in and out points
A spotting session should identify more than cue boundaries. I mark essential lines, turns in meaning and moments when the score is allowed to take foreground.
For a long narrated passage, a simple map helps: voice-led, picture-led, music-led. The balance can move between those states inside one cue. That gives the composer and editor a shared description of what the sequence needs.
I also ask how stable the narration and picture are. A shortened sentence can shift every downstream sync point. Before picture lock, modular transitions and clear restart points can be more valuable than one continuous arc that cannot survive a recut.
Sound editorial belongs in this conversation. Dense ambience, designed impacts and detailed Foley compete for attention too. Score and sound design need a shared plan for the same space.
Solve spectral conflicts in the arrangement
Intelligibility is not controlled by level alone. An instrument sitting in the voice’s presence range can blur words. Repeated attacks can interfere with consonants. Long reverbs can fill the gaps that let a sentence breathe.
I can respond through register, orchestration, density and envelope. Expressive lines can move into pauses. Sustained material can support speech while a more active layer waits for the opening.
Sidechain ducking may help in a specific delivery context. It is not a complete production plan. If the whole score drops on every syllable, the pumping can become its own distraction.
The better approach gives the re-recording mixer options without asking the cue to surrender its shape.
Deliver stems that preserve choices
A stereo master locks too many decisions together. For narration-heavy scenes, I agree useful stem groups before final delivery. Depending on the cue, those might separate rhythm, bass, harmonic beds, melodic material, solo elements and musical sound design.
Every stem starts from the same origin and runs for the same duration. File names include the cue number and version. Editorial receives a reference mix against the correct picture, not a folder of audio with an archaeological puzzle attached.
A lite mix, no-solo version or alternate ending can be useful when it answers a known risk. I do not create variants as decoration. A clear opening, a modular middle and a clean tail may offer more real flexibility than six nearly identical bounces.
If final VO or picture is still moving, I state what was used for approval. The date and version matter because a note such as “music covers the word” is meaningless if departments are listening to different cuts.
Review the sequence in three passes
First, I follow only the meaning of the narration. If a line requires effort to understand, the scene is not ready.
Second, I listen to timing. Do breaths still feel natural? Does the score push the narrator forward or make a reflective passage seem slow? Are musical accents changing the emphasis of a sentence?
Third, I check a less forgiving playback system. A balance that feels spacious on studio monitors can collapse on a television or laptop. This is not a substitute for the dub. It identifies arrangements that depend on ideal playback.
Space does not mean anonymous music
Protecting narration does not require a featureless pad. A restrained texture, harmonic change or recurring gesture can alter how the audience hears the words. Precision matters more than the number of layers.
The practical package is simple: label scratch VO, schedule final VO, spot the essential language, share plans with sound editorial and define stems before delivery. Then the final dub can make creative choices instead of performing emergency surgery.
The fader still matters. It just should not carry the entire production on its back.