A museum kiosk and a game UI share a surface — a touchscreen, short feedback sounds, a hierarchy of confirm and cancel — and it is tempting to treat them as the same problem with a different visual style. They are not. The constraints that govern a kiosk are physical and social in a way that a game’s are not, and those constraints change what a sound is allowed to be before a single waveform is chosen.
The most important of those constraints is the room itself. A kiosk does not play into a controlled mix; it plays into a gallery with HVAC noise, visitor conversation, footsteps on hard floors, and other exhibits running their own audio. The sound design has to account for that environment, and the accounting starts with a measurement rather than a preset.
Start With the Room’s Noise Floor, Because It Sets the Usable Dynamic Range
A museum gallery is not a quiet room. Studies of museum acoustics recommend an Internal Ambient Noise Level (IANL) of roughly 45 dB(A), inclusive of HVAC and visitors, as a target for exhibition spaces; guidance for communication areas within museums puts the acceptable range at 50–60 dB(A), with acoustic separation between communication and exhibition spaces at 38–44 dB(A) . One set of recommendations for unoccupied internal ambient noise is stricter, at 30 dBA or below .
Those numbers matter because they define the floor against which any kiosk sound must compete. If the gallery sits at 45 dB(A) when busy, a kiosk cue mixed at 50 dB(A) has roughly 5 dB of headroom above the noise floor. That is enough for a clear transient, and not enough for a subtle timbral detail. A sound that relies on a gentle midrange shift will disappear into the room; a sound that relies on a sharp attack and a distinct frequency band will survive.
The practical workflow is therefore: measure the occupied noise floor first, then decide the kiosk’s playback level, then design the sound to be legible at the difference between the two. The guide to reliable sound effects without studio monitors covers the cross-checking approach for rooms you cannot treat; the museum version adds a specific number to check against. A kiosk cue that only works in a quiet room is a cue that will fail on a busy Saturday.
The attack-time guide for ambient sounds is relevant in reverse here. A gallery’s ambient bed is broadband and continuous; a kiosk cue that is also broadband and continuous will fuse with it. A cue with a sharp onset and a distinct spectral signature is separable from the bed in a way that a smooth, midrange-heavy cue is not.
Neighbouring Exhibits Are a Louder Constraint Than the Room Tone
The second constraint is often more severe than the first: sound from one kiosk bleeding into the next. A gallery may have several interactive stations within a few metres of each other, each playing audio to its own visitor. If the audio spills, the result is not a richer soundscape. It is a cacophony that makes every station harder to hear and, for visitors with auditory processing differences, actively overwhelming .
The Smithsonian guidelines address this directly, recommending sound management strategies such as directional speakers or acoustic dampening to keep adjacent interactive areas from overlapping. A directional or parametric speaker can confine audio to a narrow zone — commercially available units for museum use list directivity angles of less than 15° and frequency ranges of roughly 400 Hz–10 kHz or 500 Hz–20 kHz, depending on the model .
The frequency range is the design constraint that follows from the hardware. A parametric speaker’s usable range typically starts around 400–500 Hz, which means a cue whose identity depends on content below that range will not be reproduced by the directional system. If the kiosk uses directional audio to prevent spill, the sound has to be designed to live above roughly 400 Hz — which pushes the identity into the midrange, not the sub.
The alternative is headphones, and the trade-off is different. Headphones solve spill completely and introduce a different problem: isolation. A well-known critique of audio guides is that they suppress conversation between visitors, fragmenting the shared experience into isolated channels . The Bob Dylan Center’s solution is instructive: visitors tap in-house devices against glowing symbols near exhibits to access audio through headphones, with the system delivering sound at no more than 50 ms of latency so the audio never overlaps or falls out of sync . The headphone choice solves spill; the latency constraint keeps it from feeling detached from the action.
A third option is a hybrid: directional audio for the interaction cue, headphones for the narrative content. The cue is short and needs to be local; the narrative is long and can tolerate the isolation that a headphone guide implies. The mobile app versus game UI sound guide covers the platform-expectation gap that makes this kind of layering necessary; the museum version adds the physical constraint of adjacent exhibits.
The Sensor Defines the Timing Budget, and the Budget Is Tighter Than a Game’s
A museum interaction is often triggered by proximity rather than by a button press. A visitor steps into a zone, lifts a flap, turns a crank, or approaches a screen, and the audio is supposed to respond. The delay between the physical action and the sound is not a mixing preference. It is the difference between a responsive exhibit and a broken one.
The tolerance for audio-visual integration is roughly 200 ms — the window within which a sound and a visual event are perceived as a single event . That is the outer bound, not the target. For an interaction that is supposed to feel immediate, the working target is much lower: the Bob Dylan Center specifies no more than 50 ms of latency for its headphone delivery , and a multi-user heritage installation reports an end-to-end latency of under 25 ms for its user-specific audio .
The practical implication is that the sound design has to be built around the sensor’s timing, not the other way around. A proximity sensor that polls at 10 Hz introduces up to 100 ms of delay before the system even knows the visitor is there. If the sound is supposed to start on contact, that 100 ms is already half the budget. A pressure-sensitive floor or a physical button often has lower latency than a camera-based proximity system, and the choice of sensor is therefore also a choice about what the sound is allowed to confirm.
The SFX duration guide gives the category ranges for short cues; the museum addition is that the cue must start within the sensor’s latency budget, and that a longer cue is not a fix for a late start. A 200 ms confirmation that begins 150 ms late is worse than a 50 ms confirmation that begins on time, because the former reads as lag and the latter reads as feedback.
Unattended Operation Changes What “Polished” Means
A museum kiosk runs without a sound designer present. It may run for twelve hours a day, seven days a week, for years. That changes the tolerance for failure, and it changes the vocabulary of what a sound is allowed to do.
Commercial kiosk audio hardware is specified for this environment explicitly: continuous operation of 12–24 hours daily, frequent cleaning with harsh disinfectants, occasional user abuse, and a design target of 5+ years of reliable operation using industrial-grade components such as butyl rubber cone surrounds and Kapton voice coil insulation . Consumer speakers, the same source notes, fail within 6–12 months under those conditions.
The sound-design consequence is not about hardware reliability directly. It is that a kiosk sound has to work without supervision, which means it cannot depend on a visitor learning it, and it cannot depend on a staff member adjusting the volume. A cue that is ambiguous on first hearing is a cue that will be misunderstood thousands of times before anyone reports it. A cue that is too quiet in a busy room is a cue that will be too quiet every day, because nobody will notice and turn it up.
This is the same one-pass constraint that applies to app onboarding, but with a longer time horizon. The onboarding sound guide covers the “no second play” problem; the museum version adds that the failure will be invisible to the designers and obvious to the visitors. A cue that is slightly too subtle, or slightly too similar to the adjacent kiosk’s cue, will produce confusion that nobody logs.
Accessibility Is Not an Add-On, It Is a Design Input
A public exhibition has a legal and ethical obligation to be accessible that a game does not. The Smithsonian guidelines require that exhibit materials be conveyed through multiple sensory channels, that all audio narration be provided in a printed format, and that captioning include identification of ambient or environmental sounds that are relevant to the visit .
That changes the sound design in two specific ways. First, every sound that carries information needs a non-audio equivalent, and the equivalent has to be present at the kiosk, not available on request. Second, if the kiosk uses automatic audio — a narration that starts when a visitor approaches — it falls under the WCAG requirement that any audio playing automatically for more than 3 seconds must have a mechanism to pause or stop it, and that mechanism must be available within the first three elements of the interface .
The practical sound-design decision is to keep automatic cues under 3 seconds wherever possible, so the pause/stop requirement does not apply, and to reserve longer audio for content the visitor explicitly starts. A short confirmation or transition cue can sit inside that window; a narration cannot. The accessibility design choices guide covers the broader channel-duplication principle; the museum addition is that the legal requirement turns it from a best practice into a constraint on duration and trigger.
A Constraint Checklist for the First Design Meeting
The procedure below is ordered so that the physical constraints are established before any sound is made. Each item is a question with a number attached, not a stylistic preference.
- What is the occupied noise floor of the gallery, in dB(A)? Measure it during a busy period, not an empty one. The kiosk’s playback level and the minimum signal-to-noise ratio follow from that number.
- How far is the nearest other audio source? If it is within a few metres, directional speakers or headphones are not optional. If directional speakers are used, confirm their usable frequency range — many start around 400–500 Hz — and design the cue to live above it.
- What sensor triggers the sound, and what is its latency? A camera-based proximity system, a pressure mat, and a physical button have different delay characteristics. The sound’s onset budget is the sensor’s latency plus the audio pipeline’s latency, not the other way around.
- Will the kiosk run unattended, and for how long? A cue that is ambiguous on first hearing will be ambiguous for every visitor. Design for the thousandth play, not the first.
- What is the non-audio equivalent for every informative sound? A printed transcript, a caption, a visual indicator, or a tactile element. If it does not exist at the kiosk, the sound is not accessible, regardless of how well it is mixed.
- Does any automatic audio exceed 3 seconds? If it does, a visible and keyboard-operable pause/stop control is required. If it does not, the constraint does not apply, and the cue can stay short.
A museum kiosk sound is not a smaller game sound. It is a different kind of object, shaped by a room that is louder than a game’s mix, neighbours that cannot be ignored, a sensor that sets the timing, and a public obligation that the sound be optional. The constraints above are not a style guide. They are the conditions under which the sound will either work for years without supervision or fail in a way nobody will report.