Inspiration
Every space has its own sound, but we rarely think about how much our surroundings change what we hear. A concrete stairwell can create a long, dramatic echo, while a smaller furnished room might absorb sound almost immediately.
We wanted to explore a simple question: What if you could hear how your voice would sound in different places without actually being there?
That question inspired aural, an interactive acoustic atlas of Texas A&M University. Instead of using artificial reverb presets, we wanted to capture the acoustic characteristics of real campus spaces and let anyone experience them through their own voice.
What it does
aural lets users hear an approximation of how their voice would sound in real locations across Texas A&M's campus.
Users can record their voice directly in the browser or upload an audio file, select two campus locations, and generate an A/B comparison using acoustic responses derived from our original recordings.
Beyond the voice simulator, users can explore 14 physical campus locations, compare estimated reverberation times, examine frequency-dependent decay visualizations, and listen to original flute recordings captured in different spaces.
Whether you're a musician experimenting with acoustics, a podcaster exploring recording environments, or simply curious about sound, aural makes room acoustics something you can actually hear and interact with.
How we built it
We started by leaving our computers and collecting original acoustic data across Texas A&M on November 7, 2026.
Using an iPhone's built-in microphone and the Shure MOTIV recording app, we captured three handclaps at each of 13 campus locations. We placed the phone on a stationary stand approximately five feet from the sound source and kept the recording setup consistent across locations. We also recorded flute performances in selected spaces.
Our final dataset contains 23 original WAV files, 39 individual handclaps, and nine unique flute recordings, covering 14 physical locations. Every recording was captured at 48 kHz, 24-bit, mono.
Next, we built a Python signal-processing pipeline using NumPy, SciPy, and SoundFile. We isolated individual claps, analyzed acoustic decay using Schroeder backward integration, and estimated reverberation times across different frequency bands. We then extracted approximate room impulse responses to use in our simulator.
For the website, we used HTML, CSS, JavaScript, and the Web Audio API. The application convolves the user's recording with the selected room responses to simulate each space's acoustic character. Processing happens locally in the browser, so users don't need to upload their voices to a server.
We also built interactive acoustic visualizations and deployed the application through GitHub Pages. Our complete dataset, analysis code, and reproducibility documentation are publicly available on GitHub.
Challenges we ran into
One of our biggest challenges was turning real-world recordings into usable acoustic data. Unlike recordings made in controlled laboratories, our campus recordings contained background noise, inconsistent decay quality, and differences in how individual claps were captured.
Separating the claps and estimating meaningful reverberation times required careful signal processing. We also had to distinguish reliable measurements from estimates affected by noise or insufficient decay.
Another challenge was getting the browser-based audio simulation to work consistently. We encountered a sample-rate mismatch between 44.1 kHz audio contexts and 48 kHz impulse responses, which initially prevented convolution from working. Fixing the processing pipeline allowed us to support different recording formats and generate consistent comparisons.
We also spent time removing unnecessary silence from recordings without cutting off the natural reverberation that makes each location unique.
Throughout development, we had to balance scientific accuracy with accessibility. We wanted users to understand the acoustic differences without needing a background in signal processing.
Accomplishments that we're proud of
Our biggest accomplishment is building an application where the data we collected ourselves directly powers the main experience. Without our campus recordings, aural's location-specific simulations wouldn't exist.
We're especially proud of collecting and analyzing 39 handclaps across 13 locations, turning those recordings into interactive acoustic simulations, and making the differences between spaces audible.
One particularly interesting comparison is Zachry Stairwell and Zachry Collaboration Room. Their estimated mid-frequency decay times are approximately 3.38 seconds and 0.39 seconds, respectively. Users can hear a noticeable difference when processing the same voice recording through these two spaces.
We also independently reproduced our entire acoustic analysis from the original recordings. All 13 processed room-response WAV files matched byte-for-byte, and the numerical measurements agreed within floating-point precision.
Finally, we're proud that aural is publicly accessible, requires no installation, and processes users' audio locally.
What we learned
This project taught us how much information can be extracted from something as simple as a handclap.
We gained hands-on experience with digital signal processing, impulse responses, audio convolution, frequency analysis, and reverberation estimation. More importantly, we learned how difficult it can be to transform uncontrolled real-world observations into reliable, useful data.
We also learned that a technically interesting analysis isn't enough by itself. The challenge is making that analysis meaningful to someone who doesn't understand the underlying mathematics.
Building aural pushed us to think about data collection, reproducibility, scientific limitations, user experience, and software engineering as parts of the same problem.
What's next for aural.
We want to expand aural beyond the locations we recorded during the datathon.
Our next steps include collecting recordings from more campus buildings, improving acoustic accuracy through calibrated impulse-response measurements, and adding more detailed comparisons between spaces.
We'd also like to build an interactive campus map where users can discover locations based on their acoustic characteristics, making it easier to find spaces suited to recording, practicing music, or experimenting with sound.
Long term, we envision aural as a growing library of real-world acoustic environments that lets anyone explore how physical spaces shape sound.
