Fall Detection
motion
in-house
Detect falls and sudden posture changes from motion data for always-on safety monitoring and rapid incident detection.
See it run. In one click.
The Ambient Model Zoo is a growing library of ready-to-run AI models — plug one into your dev or eval kit and watch it work in seconds. Open-source and Ambient-built, every model is tuned to run on GPX at microwatt power.
Don’t take our word for it. Watch it run.
Purpose-built edge modules. Validate your software on our evaluation kits today, and drop our SOMs directly into your final product tomorrow.
Ready to run.
Pre-trained models deploy to your kit as-is — no data collection or training run required to get started.
One click to hardware.
Flash a model to your board in a single click and see live results immediately.
Runs at microwatts.
Every model is tuned for the A-Cube architecture — full inference inside a coin-cell power budget.
Yours to shape.
Retrain any model using your own data, or bring your own model. The zoo is a starting point, not a limitation.
Proof you can hold in your hand, in the time it takes to read this page.
Demo Ready
A model for every sense.
Explore open-source and Ambient-built models across Audio, Vision, and Motion.
Gesture Glider
motion
in-house
Control gameplay through natural wrist gestures, demonstrating responsive, real-time motion inference running entirely on-device.
Voice Activity Detection and Transcription
audio
in-house
Detect when speech begins and transcribe spoken audio in real time for efficient, responsive voice applications.
Glass Break Detection
audio
in-house
Detect the distinctive acoustic signature of breaking glass for continuous, low-power home and security monitoring.
Daily Activity Recognition
motion
in-house
Classify everyday activities such as walking, sitting, standing, and movement patterns from continuous motion data.
Coming Soon
Wake Word Detection
audio
in-house
Detect a predefined wake phrase continuously at ultra-low power to activate downstream voice experiences only when needed.
Coming Soon
The whole Model Zoo
Every open-source and Ambient-built model across audio, vision, and motion - 30+ and growing.
AI models
Your eval kit, controlled from your phone.
Download ApplicationForge, pair with your kit over Bluetooth, and flash live demos in seconds — no laptop, no code.
Pair over BLE
Connect the app to your eval board in a tap.
Flash a demo instantly
Push VAD, gesture recognition, fall detection, and more to the board.
See it live
Watch inference results stream to your phone in real time.
Coming soon
on Android
on Android
Coming soon
on iOS
on iOS
Pick a model. Push a button. It’s running.
No manual porting, no glue code. ModelForge compiles any zoo model onto GPX and flashes your kit in a single step.
PICK
Choose a model from the zoo.
COMPILE
One-click ModelForge compile — quantized and mapped onto A-Cube automatically.
RUN
The model is flashed to your kit and running live. Measure the power yourself.
Works inside the standard Eclipse-based unified build.
Pre-integrated RTOS, drivers, and DSP libraries - no boilerplate.
Up to 90% of existing C code ports over.
Don’t see it? Build it.
The zoo is a starting point, not a ceiling. Bring your own TensorFlow, Keras, or ONNX model and compile it to GPX with ModelForge.
Native model import
Native TFLite import and automatic quantization to A-Cube precision.
Transfer learning
Transfer-learn from any Model Zoo model using your own data.
Architecture flexibility
If you can train it — CNN, RNN, LSTM, or GRU — you can run it.
Stop reading. Start running.
Request a kit and run any model from day one — or try it free from your phone right now.
Experience it now.
Download NuraSync and drive live demos from your phone.
Coming Soon
