Plug it in.
One cable to the wall. That is the only wire on the desk.
Get the highlights.
Finds you. Follows you.
One lens behind the band. On-device vision follows a person and keeps you in frame while it records. No face recognition or profiles.
Design
Built to sit on a desk and be looked at.
Six parts, each with one job. Scroll and the camera walks down through them.
A sphere that carries the camera.
A 1080p sensor sits behind the black band on the equator of a sphere one hundred and twenty millimetres across, in two halves that snap together. The whole sphere turns on its own axis to aim the lens. Wide enough for a room, sharp enough for a face.
An arm that raises and lowers the head.
A rigid S-shape with a single elbow. It leans back below the joint for balance and forward above it for reach, and every cable runs inside it.
Three motors. Three axes.
One at the base pans everything above it on a bearing. One at the elbow tilts the head. One in the head turns the sphere. Each is geared to move quietly and still keep up with you.
Compute on a sled.
The module stands upright in a sled inside the tower, heatsink facing the open flank. The sled slides straight out sideways with the board, its heatsink and its fan still on it, so the card and every port stay reachable without taking anything else apart.
A base that never moves.
Wide and low, like a monitor's foot. It holds the power, the speaker and the screen, and everything above it turns on a bearing as wide as the tower.
A screen on the front.
A five-inch touchscreen shows what HL-1 sees and what it is doing: resting, listening, tracking, recording. It is also where setup happens.
Tracking and motion
Three axes. One target.
Three axes of motion and on-device vision lock onto a person and keep you framed while you move and record.
Loading motion demo…
The whole upper body turns.
Everything above the base rotates on a bearing to follow you across the room, so the base stays put and the head goes where you go.
Select a motion to play it twice. Click again to replay.
Ready to play.3 axespan, tilt and aim
Person trackingnot face recognition
No profilesno memory of who it has seen
Tell it what to follow.
Say “camera” to give it a command or ask a question. It follows you, answers out loud and can describe what is in front of it.
A person
Whole-body tracking that holds on while you move.
Not an identity
No face recognition, profiles or memory of who it has seen.
A spoken command
Say “camera” and give it a command, or ask it a question. It answers out loud.
What it sees
Ask what is in front of it and its on-device vision tells you.
Voice and on-device AI
Say what you want. It does it.
A wake word, speech recognition, a language model and a voice, all running on the device. Ask it to record, give it a command, or just talk to it. No account. No cloud. No internet required.
0cloud calls for voice, vision or conversation
On-devicewake word, speech, language and voice
Works offlineWiFi is only for files and updates
Give it a personality.
Start with a preset voice and personality, then make HL-1 sound more like your choice. Explore the voice styles and accents available on the device, switch between presets, and adjust how it speaks and responds in the local settings. You can also edit the open code or install additional voices, personalities and skills locally. Your choices stay on your hardware.
Recording
Say “record”. Then forget about it.
Voice-triggered recording to on-board storage, with the space left read back to you. When you are done, pull the files to any device over WiFi over your own network, or save recordings to USB or NVMe.
Recording, and the space left.
Ask how much room is left and HL-1 tells you. Recordings stay on the robot until you choose to move or delete them. Pull files to a phone or computer over your own WiFi, or save them to USB or NVMe.
Any devicedownload from the dashboard over WiFi
Expandableover USB and NVMe
Local filesyour recordings are not sent to us
Compute and setup
A Jetson at the core.
HL-1 is designed around an NVIDIA Jetson module: GPU-class compute that keeps vision, voice and language running at the same time, on the device. A Jetson is included with Dev Kit+ and Complete. With Dev Kit, you can bring your own computer; other platforms require adaptations.
GPU on boardvision, voice and language together
1 cableto the wall, and that is all
No appno account, no sign-in
Open to builders
Ships working. Stays yours.
The included preflashed card carries the stack for Jetson-based builds, and the same image is free to download for a reflash or an update. The code is plain Python, open and editable. Add your own commands in an apps folder; updates never touch your files.
1from wafidyne import HL1
2
3cam = HL1()
4cam.follow("hand")
5cam.say("Rolling.")
6cam.record()
An SDK you can read in an afternoon.
Follow, look, say, record, listen. The whole robot is a handful of calls, and the code behind them is on GitHub for you to change. An AGENTS.md ships with it, so an AI coding agent can build against it from one prompt.
Pythonthe whole stack, readable and editable
Free updatesfor as long as you own it
Made to be built
Designed for the person who puts it together.
Assemble the enclosure, mount the components and secure them with screws where needed. The Dev Kit and Dev Kit+ let you put the hardware together yourself, with no soldering required.
It boots to its screen.
No login, app or account. The touchscreen comes up ready.
Connect your network.
Scan the on-screen code with your phone, choose WiFi on the touchscreen, or use wired networking. Say "record" to save locally. Check the remaining space, then transfer recordings over your own WiFi or save them to USB or NVMe. Nothing is deleted unless you choose to delete it.
Three ways to get one.
Choose how much of the build you want to do and whether to bring your own computer.
Bring your own computer. Build it yourself.
HL-1 Dev Kit
Bring your own Jetson Orin Nano or another computer, such as a Raspberry Pi. The mounts, software and AI models are designed for the Jetson; other computers require hardware and software adaptations.
- Compute module
- Bring your own
- Assembly
- You build it
- Calibrated and tested
- —
Compute included. Build it yourself.
HL-1 Dev Kit+
Everything in the HL-1 Dev Kit, plus the Jetson Orin Nano. Updated and boot-tested, in its own box. The full build, from the first part.
- Compute module
- Included
- Assembly
- You build it
- Calibrated and tested
- Module boot-tested
Assembled, calibrated and tested.
HL-1 Complete
The finished robot, ready on your desk the day it arrives. The same open code, without the assembly.
- Compute module
- Included
- Assembly
- Done for you
- Calibrated and tested
- Yes
Specifications.
What is in it and what it does, as designed. Software items are in development and ship as updates.
- Compute
- Bring your own computer with Dev Kit; Jetson Orin Nano Super Developer Kit is included with Dev Kit+ and Complete.
- AI
- Language, vision and voice run on the robot. Optional web lookups for chat are off by default.
- Movement
- Pan, tilt and head nod. Follows a person and keeps them in frame while recording.
- Camera
- 1080p at up to 60 fps. Clear-HDR.
- Lens
- 135° wide-angle lens.
- Included storage
- 64 GB microSD card, preloaded with the system image.
- Interface
- 5-inch touchscreen, USB camera, mic and speaker
- Construction
- 120mm sphere head. 3D-printed enclosure.
- Software
- Apache 2.0 open source. Python SDK. The included 64 GB microSD boots into cameraman mode; the image is also available to download.
- Recording
- Records locally. Pull clips over your own WiFi, or save to USB or NVMe.
- Privacy
- No account, subscription or telemetry. No face recognition, profiles or memory of who it has seen.
- Customization
- Locally installed voices, personalities and skills. Edit commands and behaviors in the documented codebase.
- Power and assembly
- One wall plug. No soldering.
12-month limited warranty
Read the coverage, exclusions and claim process for the configuration you choose.
Read the warranty14-day unopened returns
All HL-1 configurations. Buyer pays return postage and a 10% restocking fee; Complete returns need insurance and signature. Defects have separate coverage.
Return policyReal people on support
Questions before or after you buy go to the people who build the hardware. Wafidyne is a brand of Wafi Intelligence LLC.
Get supportQuestions, answered.
About HL-1 specifically. Orders, returns and warranty live on the support page.
No. Tracking, conversation, vision, voice and recording run on the robot. WiFi is used for setup, moving recordings and updates you choose to install. Optional web lookups for chat are off by default.