Audio Classifier uses your device's microphone and an on-device machine-learning model to classify sounds in real time — without ever recording, storing, or uploading your audio.
Powered by Google's YAMNet model running entirely on your device, the app analyses audio frame by frame and returns confidence scores across hundreds of sound categories. The moment a frame is processed, the audio is gone. What remains are only the numbers — labels and scores — which you can review, export, and manage entirely on your own terms.
How it works
The microphone feeds a live audio stream directly into the ML model. The model outputs a confidence score for each sound category it recognises, dozens of times per second. When scores cross your configured thresholds, an event is logged. No audio file is ever written. No recording takes place.
Presets
Configure exactly which sound categories you want to track and set the sensitivity thresholds that suit your environment. Save multiple presets and switch between them depending on what you're monitoring. Each preset stores your chosen categories, thresholds and display preferences so you can get started quickly.
Live monitoring
The monitor screen shows real-time confidence meters for every active category, highlights detected events as they happen, and tracks session statistics including total frames analysed, event count, event percentage and peak confidence. A session timer keeps track of how long you've been running.
Background operation
Start a session and put your phone away. A foreground service keeps the ML model running with a persistent notification showing that monitoring is active. Your phone's screen can stay off the entire time.
Session reports
At the end of a session, generate a summary report covering event frequency, timing patterns and confidence peaks. Reports are produced from classification scores — not audio — and saved locally as files you can export and share from within the app.
Storage and cleanup
All session data is stored in your device's private app storage. The built-in cleanup screen lets you review and delete old session logs to reclaim space whenever you need to.
Privacy by design
No audio is ever recorded. No data leaves your device. No analytics, no advertising SDKs, no crash reporters. Cloud backup is disabled. Everything this app knows about your sessions lives only on your device, under your control.

Lee Shaw

Additional Information