
Configure your robot platform and add-ons. Our team will contact you to confirm availability, delivery timeline.
Industrial humanoids are tools. Companion bots gather dust. We build character humanoids with first-tier whole-body motion: mecha, fantasy characters, future warriors. Domo, our developer platform, ships today.
LIMITED SPOTS - CURRENT BATCH NOW OPENDomo is the developer edition of that platform, built for whole-body policy learning, manipulation research, and scalable robot development from simulation to reality. Every unit in the field trains the same policy stack our characters will run on.
Up to 2 hours of continuous operation on a single charge. Fast charging reaches full capacity in just 30 minutes. Supports swappable battery packs for uninterrupted use through alternating charging cycles.
Powered by advanced language models, enabling natural conversation, task understanding, and intelligent interaction. Developers can build custom skills, policy control flows, and voice-driven robot behaviors on top of an extensible AI interface.
Built with a modular body architecture, replaceable end-effector interface, and expandable sensor system. Developers can swap between dexterous hands, grippers, and custom attachments, while adding depth cameras, LiDAR, and other perception modules for different research and application needs.
Recognizes human gestures for intuitive robot control. Use gestures to trigger actions such as auto return, navigation, walking, running, and custom interactive behaviors.
Through extensive experimentation, we found that a lightweight body, and high-torque motors together create the ideal setup for full-body control.This balance of scale and power makes it exceptionally safe and practical — perfect for laboratories, individual developers, and startups to test control algorithms at lower cost.Its fully-metal, enclosed structural design ensures high durability and deployability, making it reliable for continuous research and real-world applications.
High quality data collection is the foundation of policy training.
We provide scalable pipelines for motion generation, whole body teleoperation, and manipulation data collection-built for developers and researchers.

Convert human movement video, robot motion data to Robot format data which can be easily used in policy training

Use VR to control the robot with whole-body motion. The robot mirrors the operator’s movements, providing a fast and intuitive way to generate training data.

Collect dexterous manipulation demonstrations through intuitive leader-arm teleoperation for grasping, handling, and real-world manipulation tasks.
Hardware is where this starts, not where it ends. Three things are being built on top of the platform, they ship with the consumer line, and the reason we can build them at all is the same reason the shells come off.

Motions you call and combine instead of programming one at a time, plus a choreographer for building your own. They all run on the same policy, so a new motion can be added without rewriting the robot.

Speech and gesture as the way in. Talk to it and it responds. Wave and it comes back. It keeps small motions of its own when nobody is asking for anything, so it reads as someone in the room rather than a machine waiting for input.

Data comes back from robots in the field, policies improve on it, and the improvements go out to robots already in homes. The one you buy is the worst one you will ever own.
Rotaku doesn't ship robot models. We ship a body that characters live on. The skeleton is modular and was built from the start to change shells, so a new character is new armor and a new personality on hardware that already walks, fights and works.

Every Rotaku is an inner skeleton with standard interfaces. Armor mounts onto it and comes off it, and a different character goes on. Each plate's visual center sits off its joint axis, so the body reads as one solid piece and keeps every degree of freedom when it moves.

Modding culture only grows on machines people love. Elsewhere it does not even work: there is no character to build on, and once you change the shell the robot stops moving properly. Ours takes a new shell and still walks. That is the first time a user-made character is physically possible.

Owners swap armor, change colors and add their own marks, and the finished robot is itself the content. Creators build shells, skins and motions and sell them to owners. More creators make ownership deeper; more owners make creating worth it.
Two humanoid platforms designed for developers — a compact model for rapid experimentation, and a larger, more powerful platform for advanced full-body control and research.

Developer-focused humanoid platform for full-body control research, manipulation experiments, and custom behavior development.
A larger and more powerful humanoid platform for advanced full-body control, locomotion research, manipulation experiments, and high-impact demonstrations.
- 23 DOF
- 120 Max torque
- Approx. 35 in / 90 cm humanoid platform

- 23 DOF
- 150 Max torque
- Approx. 51 in / 130 cm humanoid hardware

more Advanced AI processing unit, more sensors, customizable

Lorem ipsum dolor sit amet consectetur dictumst amet vitae facilisi arcu odio ut habitant vestibulum.
Stories from Rotaku — our progress in humanoid robotics, physical intelligence, product development, and the mission behind building robots for the world.
Reservation priority is based on the order in which deposits are received. Because this batch has limited availability, submitting a deposit early gives you the best chance to secure a unit from the current production batch.Yes. This batch has a very limited number of available spots. Reservations are processed on a first-come, first-served basis, so we recommend submitting your deposit as early as possible to secure your position.A fully-enclosed aluminum body provides strong resistance to impact and long-term experimental wear.
After your reservation is submitted, our team will confirm availability, configuration, and delivery schedule.Once fulfillment begins, delivery typically takes 2–4 weeks.
This humanoid is purpose-built for developers, researchers, and early-stage robotics teams. It is ideal for:
• Locomotion policy training
• Full-body control
• Manipulation research
• Teleoperation
• Vision-based policy learning
• Reinforcement learning / imitation learning
Yes. An optional teleoperation kit enables scalable data collection for imitation learning and manipulation policies.
No. Domo is designed as a fully untethered humanoid platform. In the shipped version, all major wiring is enclosed inside the body shell, with onboard battery, compute, sensors, and motor control integrated into the robot. Developers can connect to Domo wirelessly, deploy and run models onboard, and access the system through SSH for development, testing, and debugging — without relying on external control cables during operation.
A fully-enclosed aluminum body provides strong resistance to impact and long-term experimental wear.