From "a noise happened" to understanding what it was
The first SilentBuddy concept detects sound. The research opportunity is a device that could recognise what it hears.
None of the AI described on this page exists today. It is future research and development. Each part is labelled so you can see what is current, what is planned and what is only a possibility.
- Currentexists or is done today
- Plannedwork we intend to do
- Potentiallonger-term possibility
The concept today: sound detection
A microphone with adjustable sensitivity listens during the quiet period. If the sound passes a set level, the device responds. At the wake time the owner has set, the light and melody come on.
The limit of this approach is that a microphone only knows that a noise happened. It can't tell a bark from a door closing.
Research: recognising the sound
AI sound recognition could allow the device to tell different sounds apart, respond more appropriately and trigger less often by mistake.
- Barking
- Whining
- Howling
- Household and environmental noise
- Different patterns of vocalisation
Longer term: recognising the dog
Further ahead, a system could potentially learn patterns specific to one dog.
- When this dog usually wakes
- Repeated wake behaviour
- Triggers that recur in the home
- Whether behaviour is changing over time
Running it on the device
The roadmap includes exploring on-device AI: recognition that runs partly or fully on SilentBuddy itself instead of sending audio to the cloud.
- FasterA response without a round trip to a server.
- More privateLess audio leaves the home.
- Works offlineThe routine doesn't depend on an internet connection.
- Lower running costLess reliance on cloud processing.
Privacy principles
A device with a microphone in a bedroom has to earn trust. These are the principles we are designing to. Specific commitments will follow once the technical design is decided.
- On-device processing where practical
- Minimal audio retention
- Privacy by design
- Clear user consent
- Secure processing
- Transparent data controls
Working in embedded AI, audio or electronics?
We're looking to bring in specialist engineering and AI expertise. The staged AI roadmap and its commercial case are on the investor page.