Synthetic test signals
Automated tests check the detection with artificially generated signals mixed with interference such as speech and music. With pure noise, the detection stays silent.
A smoke alarm with a flat battery or a charger is beeping, but you can’t find where the sound is coming from. BeepHunt remembers exactly this signal and, as you walk around, tells you whether you are getting warmer or colder.
The app asks you one question, you tap one button, and everything else runs in the background. Your phone’s microphone is all the search needs.
Every recurring tone the phone has heard.
Warmer or colder, with best value and prediction.
Turning on the spot gives a rough direction.
Optional direction finding with two devices.
Only high beeps, only low hums, or everything.
No recordings, no account, no ads.
Candidates. The app lists the recurring tones it has heard. The ones with the steadiest rhythm come first, not the loudest.
Finder. Here you see the signal strength from 0 to 100, the trend and the best value so far. The app also tells you how many seconds it will be until the next beep.
Radar sweep. If you turn round once on the spot, you get a rough direction. This feature is still experimental.
A beep is too short for you to follow it by ear. On top of that, the walls reflect the sound, so it sounds much the same all over the room.
The microphone hears every source at once. The beep first has to be separated from this mix.
With high tones, just two centimetres in a room can decide whether the sound is loud or quiet.
To hear a direction you need two separate channels, just like two ears. Many phones only provide one.
Standard measuring tools hardly help either: a sound level meter measures every noise at once, and a spectrum analyser shows the frequencies but leaves the interpretation to you.
The audio is only analysed on your device.
Offline. BeepHunt stores no recordings and uploads nothing. There is no account and there are no ads, and the microphone only runs during a search.
€2.99 once. No ads and no subscription.
Languages and system. The app is available in German and English. It runs on Android 9 or later and takes up about 5 MB.
Help in the app. BeepHunt includes an introduction and answers to common questions. You can send feedback straight from the app by email.
The first level only needs a microphone and is complete on its own. Each further level is only added if your device really supports it.
No level gets in the way of a lower one. If, for example, the connection to the second phone drops, the search simply carries on.
Some devices deliver two channels that actually come from a single microphone. The app detects this and then does not take a bearing.
BeepHunt is written in Kotlin with Jetpack Compose. The signal processing is a separate module and uses no third-party libraries.
| Module | Content |
|---|---|
| Computing core | Signal processing, detection and geometry. Because this part does not depend on Android, it can be tested quickly. |
| Recording | Microphone, ring buffer and the check of what the device can do. |
| Platform | The sensors and the connection to the second phone over your own network, without any third-party service. |
| Flow | The search survives rotating the device. |
Recording is never blocked. If the analysis can’t keep up, it skips an event rather than read distorted audio.
Automated tests check the detection with artificially generated signals mixed with interference such as speech and music. With pure noise, the detection stays silent.
In the diagnostics view the app plays the same reference tone that the tests use.
Only from this Android version onwards does the system report where the microphones sit in the device. Without that information no angle can be calculated.
BeepHunt has been on the Play Store since 2 October 2026.
First came detection, recognition, rhythm detection, tracking, the radar sweep and pairing with a second phone.
Then followed the list, the finder, the radar, the introduction and the help, in German and English.
The app was tested on other people’s devices, and all of the testers’ suggestions have been implemented.
Google has reviewed and approved the app.
BeepHunt costs €2.99 and is available in 172 countries.
The wording throughout the app has become clearer, and the display now turns with the phone when you hold it away from you.
For anyone interested, this section explains what can and cannot be worked out from a microphone signal. You don’t need it to use the app.
The longer the stretch of signal the app analyses, the more precisely it can determine the frequency, meaning the pitch, but the less precisely it can tell the timing. There is no way around this limit.
That is why BeepHunt works with three resolutions at once: a fast one for the start of a tone, a medium one for detection and a fine one for the frequency.
A fixed threshold above which a tone counts as detected never suits every environment. So BeepHunt rates every level relative to the neighbouring frequencies. The method comes from radar engineering and is called CFAR.
A guard gap around the tone stops it from raising the estimated background itself.
As you approach the source, the level changes by around 20 dB. The feature the app uses to recognise the tone must therefore not depend on loudness.
Frequency counts most, because it changes neither with distance nor with direction.
If a beep is drowned out by noise, the sequence of intervals might look like this: 2.0 · 2.0 · 4.0 · 2.0. The rhythm would then count as irregular, and the countdown to the next beep would stop working.
Such intervals are therefore folded: 4.0 becomes two beats of 2.0.
In a room, the direct sound and the reflections from the walls overlap into a fixed pattern of loud and quiet spots.
So the app only reports whether you are getting warmer or colder and remembers the best value. That way you notice when you have walked past the source.
Two microphones hear the same sound wave one shortly after the other. From this time difference the angle to the source can be calculated.
On a phone this is only unambiguous up to about 1.3 kHz, and almost every beep is higher than that. So BeepHunt also uses the envelope of the correlation, meaning the rough shape of the comparison between the two signals, to find the right peak.
Two microphones cannot tell whether the source is in front or behind, so the app shows both possibilities. Because the speed of sound depends on temperature, there is a separate setting for it.
The clocks of two phones drift apart. After just two minutes the difference amounts to a good metre of sound travel. That is why the app measures this offset first.
To do this, each phone plays a short chirp, and both record both chirps. For this measurement, sound is more precise than Wi-Fi.
Two devices can only narrow the source down to a curve; it takes a third microphone to narrow it down to an area. The app draws exactly that.
If the devices are almost in a line, the area becomes a long crescent. In that case, move one of the phones to a different spot.
| What the app does not show | Reason |
|---|---|
| A distance based on loudness | Standing waves distort the level. |
| Whether the source is in front or behind, with only two microphones | A single time difference does not fix a direction in space. |
| An exact location with only two devices | Two devices give a curve, not a location. |
| A location from devices standing in a row | The error grows strongly along the line. |
| A direction after only half a turn | Areas the phone did not cover stay empty. |
| Which device is beeping | The app only describes the tone, for example “Short beep · High”. |
| That a signal is harmless | If the tone sounds like a warning device, the app points this out in neutral terms. It never gives the all-clear. |
If you would like to know how BeepHunt is built, why certain technical decisions were made or whether the approach could work for your own use case, feel free to get in touch using this form.