Neuro N6
Vision AI. Made Simple.
Run real-time neural networks, detect objects, classify sounds, and build smarter devices.
No cloud. Low latency. Milliwatts of power.

Powered by the STM32N6
One of the most advanced microcontrollers ever brought to an Arduino compatible platform. Arm Cortex-M55 CPU paired with ST's Neural-ART NPU at 1 GHz - exceptional real-time ML on embedded hardware, with a built-in ISP, H.264 encoding, and NeoChrom GPU. A complete vision pipeline on a single board.
- Arm Cortex-M55 + Neural-ART NPU at 1 GHz
- ISP supporting cameras up to 5MP
- Hardware MJPEG and H.264 encoding
- NeoChrom GPU for image manipulation
- 64MB + 32MB OSPI RAM


One of the first Arduino boards with a dedicated NPU
The entire AI inference pipeline - camera capture, pre-processing, NPU inference, and result handling - lives inside a single Arduino sketch. If you can blink an LED, you can run a neural network.
Ohm Lab's high-level libraries abstract away all the complex STM32N6 peripheral configuration. No manual DMA setup, no register fiddling, no custom linker scripts. Just install the library and start building.
1#include <NeuroN6_app.h>2#include <OV5640_Arduino.h>3#include <PostProcess.h>4#include <Models.h>5 6NEURON6_DECLARE_MODEL(yolov8_mpe); // Declare YoloV8 Multi Pose Estimation Model7 8void setup() {9 ov5640_init(WVGA, MIRROR_FLIP_NONE); // Initialise OV5640 Camera10 DCMIPP_USB_Init(800, 480); // Initialise STM32N6 DCMIPP peripheral for Neuro Studio11 DCMIPP_NN_Init(256, 256); // Initialise STM32N6 DCMIPP peripheral for Neural Network input12 USB_feed_start(FEED_CONTINUOUS); // Start NeuroStudio feed in continuous capture mode13 NPU_Init(); // Initialise the NPU14 app_postprocess_mpe_yolo_v8_ui_init(0.4f, 0.5f); // Initialise the post processing with confidence and IOU threshold15}16 17void loop() {18 NN_feed_start(FEED_SNAPSHOT, &NN_Instance_yolov8_mpe); // Capture one frame into NN input19 NPU_Run_Inference(&NN_Instance_yolov8_mpe); // Run inference on frame20 app_postprocess_mpe_yolo_v8_ui_run(&NN_Instance_yolov8_mpe);// Post process outputs21 Display_NS_Yolo_V8_MPE(); // Build overlay info for Neuro Studio22 Send_Frames(); // Send frames and overlay data to Neuro Studio23}What will you build?
600 GOPS in a Feather form factor, drawing milliwatts. The bottleneck is your imagination.
Wildlife & trail cameras
Species detection at the edge. Classify animals without uploading thousands of frames to the cloud. Run for weeks on a battery pack.
Autonomous robots
Obstacle avoidance, object tracking, and gesture control - all on-board. No tether to a laptop, no Wi-Fi dependency.
Wearable devices
Gesture recognition, fall detection, and activity classification in a board small enough to wear. USB-C charging built in.
Smart cameras & security
Person detection that actually runs 24/7 on a battery. Trigger alerts locally - no subscription, no cloud footage.
Sound & keyword detection
Classify sounds, detect keywords, or spot anomalies in machinery noise. Inference runs in milliseconds, always listening.
Custom defect detection
Train and deploy your own inspection models. Validated internal workflow for custom model quantisation and deployment.
Pick your sensor
Neuro Vision modules connect directly to the back of the Neuro N6, allowing for rapid prototyping and flexibility in your projects.
Neuro Vision ST Cam
Neuro Vision Thermal
Neuro Vision OV5640
Neuro Vision OV5640-W

Neuro Studio
Open source. Lightweight. Ships with the board.
Connect over USB, open Neuro Studio, and see exactly what your NPU is doing in real time. Bounding boxes, class labels, confidence scores, frame rate, and inference timing - all live, no config required.
Open source - contributions welcome on GitHub

Where things stand
Everything shown is real hardware and real software. Regular updates go out to backers and the mailing list. Docs and examples land on GitHub as they're ready.
- Booting on STM32N6 production silicon
- Live video capture from camera modules
- USB streaming and direct LCD display
- Multiple neural networks running on-device
- Hardware design finalised - no further changes planned
- Full AI applications compiling and running in Arduino IDE
- Firmware upload via standard Arduino workflow
- High-level APIs for STM32N6 peripherals
- Post-processing overlays for LCD and Neuro Studio
- Custom model deployment workflow validated internally
- Live camera preview from device
- Real-time inference output visualisation
- Debug overlays and logging tools
- New features planned based on community feedback
- Open source - contributions welcome
Hardware and software,
fully open source
Everything we build for the maker community is open. The Arduino core, Neuro Studio, and hardware design files will all be published on GitHub as development progresses - so you can follow along, contribute, and start building before the board even ships.
github.com/ohmlab-ltdArduino Core
In developmentFull library support for the STM32N6 NPU, camera, and peripherals. Install via Arduino Library Manager.
Neuro Studio
In developmentOpen source desktop app for live inference visualisation, debug overlays, and session recording.
Hardware Design
Post-launchSchematics and PCB files for the Neuro N6 and all Neuro Vision camera modules.
Model Workflow
In developmentTrain, quantise, and deploy custom models. Docs and tooling will be published as the workflow matures.
Back the Neuro N6
Pre-order on Kickstarter and lock in early backer pricing. Or join the mailing list for updates, early Neuro Studio access, and Arduino library drops.
Pre-order on KickstarterJoin the mailing list for updates
No spam. Unsubscribe any time.
Building for industry instead? →