When an SFX and the music seem to fight, the first instinct is to turn one of them down. That instinct produces a mix where both are quieter and neither is clearer, because the collision was almost never a level problem. It was a frequency problem. The SFX and the music were occupying the same critical band, and no amount of volume adjustment changes which one the ear can resolve.
The exception is the case where the music was written without any space for gameplay sound — no gaps, no dynamic movement, a continuous wall of energy in the same range the SFX needs. In that case, no amount of mixing can fix the collision, because the collision was composed in. The working order is therefore: check whether the music has space, carve a band for the SFX if it does not, and only then use ducking as the runtime tool. Ducking a collision that was built into the music is treating a composition problem with a compressor.
What the Collision Actually Is: Two Sounds in the Same Critical Band
The auditory system divides the spectrum into critical bands, and within each band it resolves one dominant signal at a time. When an SFX and a music element share a critical band, the louder one masks the quieter one, and the quieter one is still present in the waveform but not perceived. The collision is not “they are both too loud.” It is “they are both in the same place.”
Enotria: The Last Song is a useful case because the game was built around music-derived sound design, which means the collision risk was structural from the start. The team tested each sound’s likely frequency range and how often it would trigger alongside others, then assigned player damage sounds to the midrange specifically so they could be distinguished without overlapping enemy damage feedback. Each weapon and spell was given a distinct pitch so that two sounds would not occupy the same frequency space[reference:0]. The mix was solved at the design stage, not at the bus fader.
The lesson for a smaller project is that the frequency assignment can be decided before the sound is made. A mix hierarchy guide covers the bus structure that makes those assignments consistent; the frequency decision sits one level below it, at the asset.
The Music Side: Write Gaps In, or You Will Be Ducking Forever
The cleanest solution to a frequency collision is to not have one. A mix where the music already has a hole in the band the SFX needs is a mix where the SFX can sit at a natural level without any ducking at all. One game audio text puts it directly: either create a sound that stands out from the music, or create a hole in the frequency spectrum of the music so the effect can be heard[reference:1].
This is easier when the music is being composed for the game and harder when it is licensed or pre-existing. For a composer, the brief is concrete: leave the midrange comparatively open during combat, avoid continuous sustained energy in the band where weapon and impact sounds will live, and use arrangement — not just volume — to create moments where the mix can breathe. A pulse that ducks itself every bar is already doing sidechain work in the composition.
For licensed music, the same principle becomes an EQ decision. Subtractive EQ on the music bus, scooped in the band the SFX occupies, is the static version of the hole. It costs some music energy, and it should be decided per track rather than applied as a blanket curve. The ducking and sidechain guide covers the dynamic version; the static scoop is the version that works when the collision is constant rather than moment-to-moment.
Ducking Is a Runtime Tool, and It Has Two Kinds
Ducking is the automatic lowering of one bus when another bus is active. The implementation is usually sidechain compression or bus volume automation, and the parameters that matter are attack, release, and depth.
A widely used set of starting values for a game ducking system is an attack of 0.1–0.3 s, a release of 0.3–0.8 s, and a duck depth around −12 dB, with a priority order of dialogue above SFX, SFX above music, and music above ambience. Those values are a starting point, not a spec: a faster attack is more transparent on a short SFX, and a slower release is less noticeable on a music bed. A common practical refinement is a fast attack around 10 ms and a slower release around 300–500 ms so the music recovers smoothly rather than pumping[reference:3].
The second kind of ducking is frequency-selective, and it is the one that solves the collision described above. Instead of lowering the whole music bus, it lowers only the band that competes with the SFX. FMOD Studio 2.03 introduced a spectral sidechain effect that follows a specific frequency range of a signal and a spectral sidechain modulator that adjusts another property based on that followed level[reference:4]. Wwise’s Meter effect combined with RTPC curves can achieve the same result, and the Baldur’s Gate 3 team describes moving from volume-based ducking — which produced what they called a chaotic “EDM-like” mix — to a frequency-based side-chaining approach that carved out specific ranges when critical sounds needed to shine[reference:5].
The difference between the two is audible. Full-band ducking makes the music feel like it is being pushed down and released; frequency-selective ducking makes a space in the music without the overall level moving. The layering guide covers the same principle from the SFX side: separate the layers by band so they do not compete. Ducking is the dynamic version of that separation.
The Level Relationship That Makes the Fight Unnecessary
A mix hierarchy that places music below SFX from the start removes most of the need for aggressive ducking. A commonly used relative-level scheme puts player SFX at −3 to −6 dB and music at −6 to −12 dB, with music described as a foundation that ducks for voice[reference:6]. The numbers are approximate, but the relationship is the point: music is not at the same level as the feedback the player needs.
The reason this matters is that a collision is much more audible when the two elements are close in level. If the music sits several dB below the SFX, a frequency overlap produces masking of the quieter element, which the player does not notice. If the two are at the same level, the overlap produces a fight that the player hears as “the sound is muddy” or “the hit does not land.” The equal loudness guide covers how to match the SFX set internally; the music relationship is the layer above it.
When the Music Should Win
Not every SFX should sit above the music. A game that uses music as a core mechanic — a rhythm game, a musical narrative, a boss fight scored as a set piece — may deliberately let the music lead and the SFX sit inside it. The difference is whether the SFX is carrying feedback the player needs to act on.
The rhythm and voice blip guide covers the case where SFX must land on the beat rather than fight it; the general principle is that a sound which is supposed to be part of the music should be mixed into the music, not over it. A sound that is supposed to confirm an input should be mixed over it. Deciding which category a sound belongs to is the mix decision that precedes every fader move.
A practical test for that decision: mute the music and play the scene. If the SFX still reads, it belongs above the music. If the scene feels empty without the music and the SFX alone is not carrying the moment, the SFX belongs inside the music, and the fight was a sign that the hierarchy was wrong rather than that the sounds were.
The sequence that avoids most collisions is therefore not a single technique. It is: decide the frequency assignment before the sounds are made, write or EQ the music with the gameplay band in mind, set a level hierarchy where feedback sits above mood, and use frequency-selective ducking only for the moments where the hierarchy has to be suspended. The sounds and the music do not have to be enemies. They have to be assigned different jobs in the same spectrum.