Real-Time Tinnitus Sound Synthesis iOS App

Every tinnitus app hands the user the same library of rain and white noise. Notch does the opposite: it synthesises the sound the user actually hears. Six sliders — pitch, width, fine tune, tremble, warmth, drift — until the phone is playing the exact thing in their head.

From that one piece of information, three things follow. A shareable sound card (PNG, PDF and a real WAV rendered by the same engine) that lets someone finally demonstrate their tinnitus to a partner or a doctor. Night masking with a one-octave notch carved exactly at their frequency, so the noise wraps around the tinnitus instead of sitting on top of it. And a guided residual-inhibition session that calibrates the stimulus, plays it, then goes completely silent and times how long the tinnitus stays away.

The audio path is one pure function — parameters to DSP coefficients to samples — deterministic, seeded, and pinned by golden-value and render-snapshot tests. It runs in a single AVAudioSourceNode render block with a lock-free triple buffer, a hard master-gain cap and a one-pole ramp on every gain change, so it never clicks. Most competitors ship looping MP3s; this generates every sample, which is why the notch can sit exactly on the user's frequency and why the exported audio is the real sound rather than an approximation.

The app has never been released. No users, no downloads, no revenue. What is being sold is a finished, tested, documented product ready to submit, not a business with a track record. docs/SETUP.md carries an explicit list of what has not been verified yet: a device run of the Silence protocol, an iOS 18 device pass, one live purchase round-trip, and an overnight audio session. None of it is development work; all of it is verification.