Three senses › By sound
By sound
It hears what no camera can see
Sound carries far past anything a lens can resolve, so it is usually the sense that notices first — and the only one that needs no hardware at all.
Strixeo listens through the phone’s own microphone, continuously, while the camera stays shut. Each window of audio is scored by YAMNet, a model trained on 521 classes of everyday sound, running on the device in well under a millisecond. Nothing is uploaded and nothing is stored: the audio is scored and discarded.
What wakes the camera is not a single loud noise but a pattern — a movement-like class scoring above a floor, more than once inside a short span. Wind in a hedge sounds the same on the twentieth window as on the first, which is exactly how it is told apart from footsteps.
Where it stops. Sound wakes the camera; it does not name the animal. Asked directly, the model calls a red fox a dog and a roe deer a turkey. That is not a bug to be tuned out — it is a model trained on doorbells and traffic being asked a question it was never built for. The photograph does the naming, always.
What we support
No hardware, ever. Sound monitoring uses the microphone already in the phone, and there is nothing to buy, plug in or charge.
| What | Where it runs | Notes |
|---|---|---|
| The phone’s own microphone | any Android 8+ | The only option, and the only one needed |
| YAMNet, 521 classes | on the device, offline | Bundled in the app; no signal required |
| External or USB microphones | — | Not supported, and not planned. The built-in microphone is the thing that makes this free. |
What we have tested
Ourselves, on real hardware. Tested means we have run it and can say what happened. Should work means the protocol matches and nobody has tried it yet. We do not merge the two.
| Tested on | Result | Notes |
|---|---|---|
| Oukitel WP61 | Works | The phone most of our field nights run on |
| Redmi Note 8T | Works | Several weeks of continuous listening |
| Recorded wildlife audio | Tuned against it | The wake rule was set from real recordings rather than reasoned, so that it could be argued with |
Sound is the sense nobody has to buy. Every phone has a microphone, so the cheapest possible setup — a handset from a drawer and nothing else — is already listening to a wood all night for the cost of the electricity. That is not a lesser version of the product. For a great many people it will be the whole of it.