The question “how long should this sound be” usually gets answered by feel, and feel is a bad guide for duration because the ear adapts to whatever length it is given. A UI click that is 400 ms will start to feel sluggish after a few hundred presses, but it will not feel wrong on the first press. The problem is not that the sound is unpleasant. It is that the sound is occupying time it does not need, and the player registers that as lag even when the input itself is responsive.
Duration is one of the few game audio parameters where published ranges are genuinely useful, because the ranges come from two independent constraints that point at the same numbers: perceptual category boundaries and engine playback behavior. A sound that runs past its category’s range stops reading as that category, and a sound that runs long enough to hold a voice slot can affect what else is allowed to play.
The table below is the starting reference. The sections after it explain where the numbers come from and where they are the wrong answer.
| Category | Typical duration | What breaks past the upper end |
|---|---|---|
| UI click / hover | 50–150 ms (hover often 80–150 ms) | Feels sluggish; the sound outlasts the visual response |
| Confirm / cancel | 100–250 ms | Delays the next menu action; competes with subsequent input |
| Error / refusal | 150–300 ms | Becomes annoying rather than informative; overlaps the retry input |
| Pickup / collect | under 300 ms | Arrives after the player has already moved on; loses the reward timing |
| Impact / hit | 200–500 ms (combat generally 200 ms–1 s) | Tail smears into the next action; weakens the attack-to-weight relationship |
| Movement / footstep | 150–500 ms | A step that outlasts the next step creates a continuous sound instead of discrete events |
| Level-up / reward | 1–2 s (fanfare up to 2–3 s) | Becomes a cutscene; delays return to play |
| Explosion / large destruction | 1–3 s, with tail | Tail occupies voice slots and masks later events |
| Ambient loop | 30 s–2 min (seamless) | Shorter loops become recognisable; longer loops waste memory without perceptual gain |
Where the UI Numbers Come From: the 100 ms Loudness Boundary
The UI ranges are the tightest of any category, and the reason is not stylistic. Psychoacoustic research on loudness and duration shows that for sounds shorter than roughly 100 ms, loudness increases with duration up to that point; beyond about 100 ms, loudness is largely determined by level rather than duration . In practical terms, a click that is 60 ms and a click that is 140 ms can be matched in perceived loudness, but the 140 ms click occupies more than twice the time, and that extra time is heard as the sound lingering.
The category-specific ranges in published guidance reflect this. Morphic’s cue set puts interface clicks under 100 ms, confirms at 100–250 ms, errors at 150–300 ms, and pickups under 300 ms . A separate UI-sound breakdown places taps and clicks at 80–150 ms, hovers at 120–200 ms, and notifications at 200–600 ms . The numbers differ slightly because they are describing slightly different event types, but they agree on the boundary: a click that crosses roughly 150 ms starts to feel like a different kind of event. The game UI sound design guide covers the hierarchy between these categories in more detail.
There is a second, practical reason UI sounds stay short: debouncing. A typical UI debounce interval is around 150 ms, meaning the game will not play another click sound within that window even if the player presses again . A click sound longer than the debounce interval will still be sounding when the next click is suppressed, which reads as a missed input rather than a fast double-press.
Impacts and Footsteps: the Range Is Set by the Next Event, Not the Sound
Impact and movement sounds are longer than UI sounds because they are allowed to carry a tail, but the upper end of the range is not about the sound’s own character. It is about how soon the next sound of the same type will arrive.
For combat hits, published guidance puts the range at 200 ms–1 s, with a tighter 200–500 ms for a single hit . For movement, the range is 150–500 ms, covering footsteps, jumps, landings, and dashes . The overlap is deliberate: a footstep on gravel and a light melee hit can occupy the same duration window because both are discrete events that need to be readable before the next one begins. A footstep that runs 800 ms will still be sounding when the next step lands during a sprint, and the two sounds merge into a continuous texture instead of a sequence of steps. The footstep surface guide covers how surface-specific decay interacts with this constraint.
The engine layer matters here in a way it does not for UI sounds. Wwise’s Playback Limit system caps how many voice instances a sound can occupy at once, with options at the per-game-object and global level . A footstep that runs long enough to overlap with the next step consumes two voices instead of one; a combat sound with a long tail does the same. On a scene with many simultaneous sounds, the duration of each one determines how quickly the voice budget is exhausted, and voice exhaustion is audible as sounds being cut off or failing to play. The sound prioritization and voice limiting guide covers how to set those limits so duration decisions do not turn into missing sounds.
Explosions and Large Impacts: Duration Is Mostly Tail
An explosion is the category where the “duration” number is least useful on its own, because the audible event is divided into an attack (the initial blast) and a tail (the decay and reverb). The attack of a large explosion is typically on the order of 1–2 seconds before the decay becomes the dominant part, and the tail can extend several seconds beyond that . The explosion sound effect guide covers how noise shaping and filtering determine that decay character.
The practical decision is not “how long should the explosion be” but “how long can the tail be before it costs more than it adds.” A 5-second explosion tail in a scene with frequent combat will keep voices occupied and mask the next event. The same tail in a scripted destruction sequence with nothing following it is fine. This is why explosion duration is often a per-context decision rather than a per-asset decision: the same source file can be used with a shortened tail in gameplay and a full tail in a cutscene.
Ambient Loops: the 30-Second-to-2-Minute Convention and Why It Exists
Ambient beds follow a different logic because they are not events. A one-shot sound’s duration is bounded by the next event; a loop’s duration is bounded by how long it takes the listener to memorise the pattern.
A commonly cited working range for environmental ambience (wind, rain, room tone) is 30 seconds to 2 minutes. The lower bound is perceptual: a loop shorter than about 30 seconds becomes recognisable as a repeating unit, and once the listener has memorised the pattern, the loop point and the recurring events inside it become audible as structure rather than atmosphere . The upper bound is practical: a longer loop costs more memory and streaming bandwidth without a proportional perceptual gain, because the listener has already stopped tracking the pattern by the time the loop point arrives. This is why many commercial ambient packs ship loops in the 30-second to 2-minute range .
The exact number depends on the density of identifiable events inside the loop. A sparse drone can loop at 30 seconds without becoming recognisable; a dense forest bed with distinct bird calls may need closer to 2 minutes before the calls stop forming a pattern. The ambient background sound effects guide covers how layer density interacts with loop length.
When the Range Is the Wrong Answer
The table is a default, not a rule, and there are three situations where the default range will produce the wrong sound.
Weapon tails and distant reflections. A gunshot’s dry attack is short, but the reflection tail in a large space is not. If the engine’s reverb is applied as a separate send, the dry sound can stay in the 200–500 ms range while the tail runs several seconds. If the reverb is baked into the file, the file has to be long enough to contain the tail, and that changes the duration decision. The occlusion and obstruction guide covers how distance and obstruction extend the perceived tail independently of the file length.
Intentional overlap. Some categories are designed to overlap. A sustained energy weapon or a continuous spell channel is not a one-shot with a long duration; it is a looping or sustained sound, and the one-shot duration ranges do not apply. The same is true for any sound that is meant to be crossfaded with itself rather than allowed to decay.
UI sounds in slow, deliberate interfaces. A horror game’s menu may deliberately use a longer confirm sound than a fast-paced action game, because the interface’s pacing is different. The 100–250 ms range assumes the player is moving through menus quickly; if the design intent is weight and hesitation, a longer sound can be correct. The range is a baseline for responsiveness, not a cap on expression.
A Short Procedure for Settling Duration on a New Sound
The procedure below is ordered so that the fastest checks come first.
- Identify the category and start at the upper end of its range. A UI click starts at 100–150 ms, a confirm at 250 ms, an impact at 500 ms. Starting long and trimming is easier than starting short and extending, because extending a sound usually means adding silence or regenerating, while trimming is a fade.
- Trigger the sound at the rate the game will actually produce it. A footstep at sprint speed, a UI click at menu-scrolling speed, a hit at the attack-rate of the fastest weapon. If the sound is still decaying when the next instance begins, it is too long for that rate.
- Check the voice budget impact. If the sound has a tail that overlaps with itself, it consumes two voice slots instead of one during the overlap. On a scene with many concurrent sounds, shorten the tail or route it to a bus with its own limit.
- Trim leading silence before anything else. A sound that starts with 20 ms of silence will feel late even if its total duration is correct. The fix is to trim the front, not to shorten the tail.
- If the sound still feels wrong at the category range, check whether the category is right. A sound that needs 800 ms to read as a hit may not be a hit problem; it may be a weapon-weight problem that belongs in a different category, or a mechanic problem that no sound duration will fix.
The prototype versus polish guide covers the broader question of when a duration decision is worth revisiting. If a sound has been through several duration passes and still does not sit right, the problem is usually upstream of duration — in the mechanic, the mix, or the category assignment — and no further trimming will solve it.