2Loud is starting as a Chrome extension. But the thing we're really trying to build is an open dataset of how sound affects people.
People have real reactions to specific sounds every day: a dog barking at a TV doorbell, a baby startled awake by a gunshot in a show, someone flinching at a scream in a movie.
These reactions are data — data that could make your world a little more peaceful.
What if we could learn from it?
On the surface, 2Loud is a Chrome extension that ducks the volume on sounds you don't want to hear. Under the surface, the hope is that every time someone marks a sound as bothersome — or confirms that our AI got it right — they'd be helping build an open dataset.
An open, anonymized dataset of human-verified sound classifications in real media content.
Neuroscientists and accessibility researchers could get real-world data on sound sensitivity they can't get from lab studies.
Audio classification models could get human-verified training data grounded in actual media, not synthetic benchmarks.
Hearing aid companies and assistive tech developers could get signal on which sounds cause real problems in real environments.
Before any of that matters — your dog stays calm, your baby stays asleep, and you get to just watch something without bracing for the next loud moment.
Everything is designed to run locally in your browser. No audio ever leaves your device. The only data that would travel is anonymous, event-level feedback: this sound, at this timestamp, was flagged by a user.
We're choosing not to collect names, emails, demographics, or anything that could identify you — because the science doesn't need it, and neither do we.
The raw timestamp data — the most valuable part — will stay protected. Aggregate patterns will be shared openly with the research community. That's the deal, and we intend to keep it.
We're building 2Loud as a 501(c)(3) because the incentives matter.
We could build this as a startup. The data alone — behavioral patterns tied to sound sensitivity across every streaming platform — would make a compelling pitch deck. Segment it by condition, sell it to advertisers, license it to platforms. VCs would take that meeting.
We don't want to build that company. A for-profit version would eventually face pressure to monetize the data, expand collection, or compromise on privacy. We want to make those decisions irreversible from day one.
Being a non-profit means we can keep the dataset open, resist the pull to track users, and measure success the way it should be measured: dogs that stop barking, babies that stay asleep, people who can finally just watch a movie.
This is public because we want it to be. The idea is here. The vision is here. The ethical framework is here.
If you want to build something adjacent, go ahead. If you want to help build this specific thing, reach out. If you want to tell us this already exists and we missed it, please do.
We're not protecting this with secrecy. We're protecting it by being the ones who build it right.
Join the waitlist