Jetson Orin Nano Super: What to Add & What to Build Next
A Hi, Bot field guide to the accessories actually worth buying for your Jetson Orin Nano Super, plus six projects from beginner to ambitious — each with its own parts list and real prices.
June 17, 2026 · Hi, Bot
A Hi, Bot field guide — companion to the getting-started walkthrough.
You finished the setup guide — the kit boots, JetPack is flashed, and you've run your first model. This is the companion sheet: the accessories actually worth buying, and six builds from beginner to ambitious. Every project lists its own parts, with real prices where the part is sold by Seeed Studio.
The board, at a glance: 67 TOPS (INT8) · 102 GB/s memory bandwidth · 8 GB LPDDR5 · 1024 CUDA cores · 7–25 W · $249 dev kit.
Start here: what's in the box vs. what you add
The dev kit is a complete computer, but a few things are intentionally left out so you can size them to your project.
✓ In the box
- Reference carrier board
- Orin Nano 8GB module, pre-mounted with an active heatsink + fan
- Power adapter
Cooling is already handled on the dev kit — you don't need to buy a fan unless you're building from a bare module.
＋ You add
- Storage — nothing is included; the board has a microSD slot and an NVMe M.2 slot
- Networking — no wireless card; use Ethernet, a USB Wi-Fi dongle, or an M.2 E-key Wi-Fi module
- A camera — for any vision or robotics build
- Monitor + HDMI/DP cable + USB keyboard/mouse — for first setup (go headless over SSH afterward)
Core add-ons worth buying first
If you buy nothing else, buy storage. AI models are large and the microSD slot will bottleneck everything; an NVMe SSD is the single biggest quality-of-life upgrade.
Tier 1 — get these to start
| Add-on | Why | Source | Price |
|---|---|---|---|
| NVMe M.2 SSD | 10–100× faster than microSD, room for many models. 256GB to start; 512GB–1TB if you'll keep recordings. Spec: M.2 2280, NVMe. | Any reputable brand | ~$30–80 |
| Camera (CSI) | Plugs into the board's MIPI-CSI connector. IMX219 is the classic low-cost choice. | Seeed · 102110719 | $15.90 |
| Networking | Ethernet is simplest; USB Wi-Fi dongle is the no-fuss wireless option; M.2 E-key card is cleaner. | cable / dongle / M.2 card | ~$0–20 |
Tier 2 — add when a project calls for it
| Add-on | Why | Source | Price |
|---|---|---|---|
| USB mic + speaker | Needed for any voice project. A USB speakerphone is easiest. | Any reputable brand | ~$20–40 |
| High-Quality Camera (IMX477) | 12.3MP, interchangeable lenses — when image quality matters. | Seeed · 114992442 | $59.90 |
| Active heatsink + fan (spare) | Only for bare modules / fanless carriers — dev kit already has cooling. | Seeed · 110991904 | $19.90 |
| USB webcam | Plug-and-play alternative to CSI; works in every project below. | Any UVC webcam | ~$25 |
Six projects to try
Ordered roughly easiest to hardest. Each "added cost" is on top of the dev kit and the Tier-1 add-ons you've already bought.
Project 01 — Offline AI chatbot
Beginner · ~1 hr · added cost $0
Run a real language model entirely on the board — no internet, no API keys. The natural next step after the ollama / phi demo in the setup guide: swap in different models and talk to them locally.
You'll learn: running quantized LLMs locally, picking model sizes that fit 8GB, Docker, basic prompting.
Guides: NVIDIA Jetson AI Lab · dusty-nv/jetson-containers
| Item | Why | Source | Price |
|---|---|---|---|
| NVMe M.2 SSD | Models are large; SSD keeps load times sane | Tier-1 add-on (already owned) | — |
| Added cost beyond your core kit | $0 |
Project 02 — Live object-detection camera
Beginner → Intermediate · 2–3 hrs · added cost $16–35
Point a camera at the room and watch the board label people, cups, laptops in real time with live bounding boxes. NVIDIA's "Hello AI World" runs a pre-trained detector on your feed in a few commands.
You'll learn: real-time computer vision, CSI camera streaming, confidence thresholds, reading frame rates.
Guides: jetson-inference (Hello AI World) · Jetson AI Lab community projects
| Item | Why | Source | Price |
|---|---|---|---|
| IMX219 CSI camera | The feed the detector runs on | Seeed · 102110719 | $15.90 |
| Tripod / mount (optional) | Aim it steadily | Any brand | ~$8 |
| NVMe SSD | Holds model and demo data | Tier-1 add-on (already owned) | — |
| Added cost beyond your core kit | $16–24 |
Project 03 — Offline voice assistant
Intermediate · 3–5 hrs · added cost $20–40
Chain three models into a talking assistant that never touches the cloud: speech-to-text → local LLM → text-to-speech. Ask it questions out loud and hear it answer — all on the board.
You'll learn: an end-to-end audio pipeline, ASR (e.g. Whisper), TTS (e.g. Piper), chaining models.
Guides: Jetson AI Lab (speech/ASR) · jetson-containers (whisper, piper)
| Item | Why | Source | Price |
|---|---|---|---|
| USB mic + speaker (or speakerphone) | Ears and mouth | Any reputable brand | $20–40 |
| NVMe SSD | Three models is a lot of storage | Tier-1 add-on (already owned) | — |
| Added cost beyond your core kit | $20–40 |
Project 04 — Smart security cam / mini-NVR
Intermediate · 4–6 hrs · added cost $60–140
Turn the board into an always-on camera that records or pings you only when it sees a person — not every shadow. Frigate is a popular self-hosted NVR that pairs detection with recording and runs well on Jetson.
You'll learn: continuous video pipelines, event-triggered recording, motion vs. object detection, storage planning.
Guides: Frigate NVR docs · jetson-inference (detectnet)
| Item | Why | Source | Price |
|---|---|---|---|
| Camera — USB webcam or IP camera | The feed being watched (IP cam for outdoors) | Any brand · or Seeed IMX219 ($15.90) | $16–60 |
| Larger NVMe SSD (512GB–1TB) | Continuous recording fills space fast | Any reputable brand | $45–80 |
| Added cost beyond your core kit | $60–140 |
Project 05 — Robotics: control an arm with an AI agent
Advanced · a weekend+ · added cost $150–280
Give natural-language commands and watch a robotic arm carry them out. Seeed's OpenClaw framework is the agent brain; LeRobot drives the motors of an open-source SO-ARM101 arm; a camera handles perception — understanding, then physical action.
Honest scope note: Seeed's full arm-control demos run on larger Jetsons (e.g. Thor). On the 8GB Orin Nano, use a lighter local model — Seeed's Jetson-Claw starter for Orin Nano 8GB is built for exactly this constraint — or offload the heavy reasoning. The arm's motor control is light; the LLM is the limiting factor.
You'll learn: agent frameworks, motor calibration, connecting perception to action, matching model size to an 8GB board.
Guides: awesome-openclaw-hardware-projects · OpenClaw hardware
| Item | Why | Source | Price |
|---|---|---|---|
| SO-ARM101 follower arm kit (LeRobot-compatible) | The arm that moves; price varies by servo grade | Open-source kit · varies by vendor | $120–250 |
| USB webcam | Perception for the agent | Any UVC webcam | ~$25 |
| NVMe SSD | Agent framework + local model | Tier-1 add-on (already owned) | — |
| Added cost beyond your core kit | $145–275 |
Lighter alternative: a wheeled JetBot rover. Seeed sells a JetBot Smart Car Kit (often out of stock), though it targets the older Jetson Nano — check compatibility before buying.
Project 06 — Train your own image classifier
Intermediate · 3–5 hrs · added cost $0–16
Teach the board to recognize your things — pets, parts bin, hand gestures. Collect photos with the camera, train (on the Jetson or a free Colab GPU), then run your custom classifier live. This is where it stops being someone else's demo and starts being yours.
You'll learn: building a dataset, transfer learning, exporting to ONNX/TensorRT, deploying a model you trained.
Guides: jetson-inference: transfer learning · Google Colab
| Item | Why | Source | Price |
|---|---|---|---|
| Camera (CSI or USB) | Collect your own training images | Seeed IMX219 · 102110719 · or any webcam | $0–16 |
| NVMe SSD | Datasets and checkpoints | Tier-1 add-on (already owned) | — |
| Added cost beyond your core kit | $0–16 |
Keep going
- NVIDIA Jetson AI Lab — curated generative-AI tutorials for Jetson
- jetson-containers — pre-built containers for nearly every model
- jetson-inference — the classic vision tutorials
- NVIDIA Jetson forums — friendly, active, someone's hit your bug before
- awesome-openclaw-hardware-projects — robotics & agent builds
- Seeed Studio Jetson catalog — cameras, carriers, cases, kits
→ Back to the getting-started guide
About these prices: board specs and the $249 price are from NVIDIA's Jetson Orin Nano Super Developer Kit page. Seeed Studio SKUs, prices, and stock reflect a recent catalog snapshot and can change — always confirm on the product page before buying. Items marked "any reputable brand" are not sold by Seeed; the listed range is a buying guide, not a quote.
› Keep reading
Get the next dispatch.
Occasional, never spammy.