Identify the mystery
sounds in your home
— privately.
That hum, buzz, drip, or rattle you can't place? EchoFind classifies it in real time using on-device machine learning. No cloud. No account. No audio ever leaves your device.
25 sound classes. One tap.
EchoFind draws from a 25-class taxonomy covering every kind of sound a home throws at you. A fine-tuned YAMNet model runs entirely on your device — Core ML on iOS, TFLite on Android.
Because mystery sounds deserve answers, not expensive technician visits.
Everyone has heard an unidentifiable noise in their home and wondered: "What IS that — and where's it coming from?"
The usual options are bad: pay for a technician, or post a phone recording to a public forum — leaking your private home audio to strangers.
EchoFind answers the question instantly and privately. Raw audio never leaves your device. There is no account, no cloud upload, no surveillance. Classification happens entirely on-device. The only thing that can ever leave is an optional evidence package you explicitly choose to export — and in all-party-consent states it strips human speech above 1 kHz first.
Privacy is the core differentiator.
Three steps. No setup.
Tap to listen
Open EchoFind and tap the glowing record button. It starts listening entirely on-device — no connection needed.
Hunt the sound
Walk toward where the sound is loudest. EchoFind classifies what it hears in real time, showing the top match and a confidence percentage.
Review & save
See your ranked results, save the session, and optionally export a shareable evidence package for a landlord or repair tech. No account needed.
Everything you need. Nothing you don't.
On-device ML
YAMNet fine-tuned model runs entirely on your phone. Core ML on iOS, TFLite on Android. No server round-trip.
Real-time classification
Continuous inference as you move through a room. See the top sound class and confidence score update live.
Privacy-first
Raw audio never leaves your device. No mic data in transit, ever. The only export is one you manually trigger.
Source heatmap
Walk-path heat map that localizes the sound source to room level (~1 m). Powered by your movement pattern over a session.
Evidence export
Package your session — classification log, timestamps, confidence scores — to share with a landlord, repair tech, or insurer.
25-sound taxonomy
Seven categories: Mechanical, Electrical, Plumbing, HVAC, Biological, Human, Ambient. Trained specifically on residential environments.
Your audio stays yours. Full stop.
- On-device inference only — the model never talks to a server
- No account, no sign-in — nothing to breach
- Raw audio never uploaded — not to us, not to anyone
- Explicit-export-only — you decide if any data leaves the app
- Speech-stripping — exports suppress voice above 1 kHz in all-party-consent jurisdictions
Questions answered.
What sounds can it identify?
EchoFind covers 25 classes across 7 categories: Mechanical (motors, fans, vibration, rattles), Electrical (transformer hum, buzzing, static), Plumbing (drips, running water, pipe knock), HVAC (airflow, duct resonance, compressor), Biological (insects, animals, pest activity), Human (footsteps, voices — flagged but not transcribed), and Ambient (traffic bleed, weather, undefined background). Results show the top-match class name and a confidence percentage.
Does it record or upload my audio?
No. EchoFind classifies sound in real time using on-device inference — audio frames are processed locally and discarded. Nothing is recorded to a file unless you explicitly save a session, and even then that file stays on your device. There is no backend, no audio upload path, and no telemetry. The only way audio-related data leaves the app is if you manually trigger an evidence export.
Is it really free?
Yes — EchoFind is free with no ads, no in-app purchases, and no freemium gate. Future premium tiers (e.g. for the source-heatmap feature) may be added, but the core classification functionality will remain free.
What platforms does it support?
iOS 16 or later at launch, running on iPhone. Android 12+ support is planned for a subsequent release. There is no web or desktop version — on-device ML requires native platform access to the microphone and NPU/ANE.
When can I get it?
EchoFind is currently in development. Use the form below to join the notify list and we'll email you when it launches on the App Store. No spam — one email when it's ready.
Do I need an account?
No account, sign-in, or subscription required. The app works fully offline. There is no server to log into and no credentials to manage or protect.
Get notified when
EchoFind launches.
One email. No spam. Unsubscribe any time.