1 in 7

adults worldwide lives with tinnitus. For most, there is no cure — only the search for relief.

A love story, actually

Every app has an origin story. This one is a love story.

My husband has tinnitus. For him, silence isn't silent: quiet evenings, the moment before sleep, an empty room — there's always a ringing that only he can hear, and it never switches off. I watched the person I love hunt for relief everywhere: conventional medicine, acupuncture, every masking-sound playlist the internet had to offer.

Some nights he pretended something worked, so I would worry less.
Most nights, he couldn't even fake it.

I couldn't make the ringing stop. So I did what felt entirely proportionate at the time: I enrolled in a Master's in Artificial Intelligence. (He probably hoped for a more tender, affectionate wife to hold his hand through the hard nights. He's getting a thesis. He calls it weaponized autism; I call it love.) The question driving it all: could a machine learn which sound gives one specific person their quiet back? Tinnitus is deeply personal — the masker that saves one person's night does nothing for the next. That's exactly the kind of problem machine learning was made for.

MyTinnitusMask is my thesis on paper, but really it's a promise. Because it was built out of love rather than for profit, it is free, forever — no subscription, no ads, no account. Dedicated to all of you who live every day with a sound nobody else can hear, and have felt like nobody cared.

You are not alone.

What does tinnitus sound like?

If you have tinnitus, you already know — this demo is for the people who love you. Send them this page and let them hold the button.

A quiet metallic static that swells and fades in slow waves — tuned to match his clinical audiogram. It stops the instant you let go.
He doesn't get to let go.

How it works

1 · Match your tinnitus

A guided psychoacoustic session finds your tinnitus pitch (binary search over frequency), its loudness, and the minimum level of noise that masks it — the same measures used in audiology clinics, adapted for AirPods.

Psychoacoustics Frequency Matching MML

2 · Compare candidate sounds

Each session plays two candidate maskers, A/B style, and you rate the relief each one gives. Comparing two sounds in the same session filters out the noise of good and bad days.

A/B Sessions Relief Ratings

3 · The optimizer learns

A Bayesian bandit algorithm (Thompson sampling) runs entirely on your device, updating its beliefs after every comparison and converging on the masker parameters that work best for you.

Bayesian Optimization On-Device ML

4 · Listen and relax

Once rated, the winning sound keeps playing on a timer — in the background, with the screen locked — while you read, work, or fall asleep. Progress is tracked with the clinically validated THI questionnaire.

Background Audio THI Tracking

Privacy by design

Everything stays on your device: your tinnitus profile, your sessions, your questionnaire scores. There is no server, no account, and no analytics.

If you choose to contribute to the research

Contributing a session is strictly opt-in, one session at a time.
The exact data is shown on screen before anything is sent — no hidden fields.
Contributions are truly anonymous: no name, no email, no device ID, nothing that can link a row of data back to you.
The dataset, methods, and analysis protocols will be published openly (CC0) alongside the thesis.

Project status

MyTinnitusMask app icon — a jagged wave settling into a calm one

The app is feature-complete for its first study. The planned path:

Now — in beta on TestFlight, alongside hardware validation (AirPods Pro frequency characterization).
Next — ethics committee submission, then a small clinical pilot.
Later — public release. Participation in the research opens once ethics approval is in place.

Interested in the project — as a participant, clinician, or researcher? Reach out and I'll get in touch when the study opens.