—
0:00.0 / 0:00.0
Settings changed — generate again
Every space
has a sound.
Discover how your voice changes across Texas A&M. Record a sound, choose a space, and hear its acoustic signature.
Built from handclap and flute recordings our team made on campus.
The Studio
Put your voice in another room.
Record or upload a short clip, then hear it through acoustic responses captured with handclaps in two real campus spaces. The result is an approximation, processed entirely in your browser.
- 01Capture
- 02Select
- 03Listen
01
Capture
Your recording will appear here. Drop an audio file anywhere on this panel.
Original
0:00.0 / 0:00.0
02
Select
Dry voiceFull room
Add a recording to begin.
03
Listen
Switching between A and B keeps your place in the clip, so you can compare the same moment.
—
0:00.0 / 0:00.0
Settings changed — generate again
| Measured property | A | B | What it means |
|---|
Estimates from three handclaps recorded in each space. A dash means no reliable value could be fitted. See methodology.
Explore
The spaces.
Every location below was recorded by our team on the Texas A&M campus.
- Loading recorded spaces…
The Science
The science of sound.
When sound stops, a room keeps ringing. How long that takes, and which frequencies linger, is what makes a stairwell sound different from a meeting room.
Technical details
- T20
- Slope of the Schroeder decay curve fitted between −5 and −25 dB, extrapolated to a 60 dB decay. The headline value averages the 500 and 1000 Hz octave bands.
- T30
- Same method using −5 to −35 dB, reported only where the recording had enough range above the noise.
- EDT
- Early decay time, from the −1 to −10 dB portion of the curve, extrapolated to 60 dB.
- C50 / D50
- Ratio and share of energy arriving in the first 50 ms versus later. Higher values usually mean clearer speech.
- Filters
- Broadband 100–8000 Hz third-order Butterworth; octave bands at 125–4000 Hz, fourth-order Butterworth.
- Fit rules
- Fits with R² below 0.84, too few points or less than 45 ms of span are rejected. The median of the valid claps is reported. Rejected fits stay blank and are never set to zero.
Field recordings
Field recordings.
A flute played in the actual rooms and captured on site. These are real recordings, not simulations, and only the spaces listed here have one. Silence before and after the music is skipped for listening; the stored files are unchanged. They were recorded with the same phone setup as the claps, but melody and dynamics were not standardised, so treat them as references, not measurements.
Methodology
From a handclap to a room.
How the recordings were made, what we measured from them, and where the numbers stop being exact.
01Why we recorded
We wanted to answer a simple question: how differently would the same voice sound in different places at Texas A&M? Rather than borrowing published room data, the team went to campus spaces and recorded them directly. Singers, musicians, podcasters and students can use the result to preview the character of a space before going there, or to hear what reverberation does to speech, as long as they treat it as an estimate.
02How the claps were recorded
On November 7, 2026, we recorded three handclaps in each of 13 spaces (39 claps, all identified by the analysis) on an Apple iPhone's built-in microphone using the Shure MOTIV app: uncompressed 48 kHz, 24-bit, mono WAV. The phone stood still on a stand about 5 ft (1.5 m) from the sound source, in roughly the same arrangement everywhere, with the same settings throughout. We also recorded a flute in nine spaces with the same setup. One flute-only recording, filed as Blocker Stairwell, was made in the same stairwell as the Blocker Staircase clap recording, so the two are separate samples of one place; 14 campus locations in all. Exact positions, room dimensions and recording times were not recorded, so we don't report them.
03How the audio was analysed
Each clap was located, cropped from 8 ms before the peak, and filtered into a broad band and six octave bands. Background noise measured just before each clap was subtracted, and each response was cut off where it fell back into that noise.
04Schroeder backward integration
Integrating the squared response backwards from its end gives a smooth energy-decay curve. A straight line fitted to that curve gives the decay rate, which we extrapolate to the time a sound would take to fall by 60 dB.
05Frequency-dependent decay
Repeating the fit in each octave band shows whether bass or treble lingers longer. Where a band's decay could not be fitted reliably, the value is left blank.
06What the simulation does
For each space, one clap with a valid fit was cropped to 2.2 s, gently noise-gated, faded and energy-normalised. In the Studio, silence before and after your clip is trimmed (pauses inside it are kept), then the clip is convolved with that response in your browser and mixed with the dry signal by the room-effect slider. Each result keeps the room's full decay after your last sound, and both share one gain so their loudness difference is preserved.
07Limitations
A handclap is not a calibrated impulse response. It has its own sound, radiates unevenly, and was recorded on a phone microphone, which is not calibrated and may apply automatic gain; keeping the settings fixed doesn't rule that out. Positions were kept roughly consistent but not measured. The simulation conveys a space's character but does not reproduce it exactly.
08Estimates, not certified measurements
Formal reverberation tests (ISO 3382) use a calibrated source, controlled positions and multiple measurement points. Our T20, T30 and EDT values are exploratory estimates from handclaps, useful for comparing these spaces with each other, not as certified acoustic ratings.