Let Intelligence Flow.

Capabilities Product Details
Scroll Down
About Us

Let Intelligence Flow
Across Embodiments, Industries, and Time.

Litchi Lab builds the intelligent infrastructure of the Physical AI era, breaking down the barriers between scenarios, data, models, and hardware so intelligence can flow efficiently across different hardware embodiments and application environments. Through an open, unified intelligence foundation, the company lets different robot platforms share autonomous navigation and manipulation capabilities, while helping customers in manufacturing, logistics, healthcare, and services move beyond single-hardware lock-in and flexibly build Physical AI solutions for the real world.

By connecting more embodiments and serving more industries, Litchi Lab is building an open, continuously evolving Physical AI ecosystem where intelligence keeps migrating, accumulating, and growing in the physical world.

For Hardware Makers

Universal Operating System

A universal operating system for the Physical AI era. The same intelligence stack adapts across embodiments and freely drives different hardware categories — any model, any form, plug in and run. Open platform interfaces and no lock-in let hardware makers focus on embodiment innovation.

"Let intelligence run free across every form."

For End Customers

Flexible Solutions

Flexible physical intelligence solutions for manufacturing, logistics, healthcare, services, and every industry in between. From data collection to policy deployment, tailor as needed and adapt to each scenario, so physical intelligence is truly deployed and truly usable.

"Let intelligence serve every industry."

Core Capabilities

Let Intelligence Cross Forms, Enter Scenarios, and Keep Evolving

Let Intelligence Flow Across Embodiments, Industries and Time

01
Intelligence Flows Across Every Form
Intelligence is not bound to any single robot
The same intelligence stack adapts freely to robotic arms, dexterous hands, humanoid robots, and other embodiments, decoupling intelligence from hardware. Different robots share one intelligence system, and each body unlocks different value.
02
Intelligence Flows Across Every Industry
Intelligence is not limited to any single scenario
From manufacturing and logistics to healthcare and services, one intelligence system migrates across industries without being rebuilt for every scenario, bringing Physical AI into the real world faster.
03
Intelligence Flows Through Continuous Evolution
Intelligence is more than a single training run
Every deployment, every interaction, and every data loop enriches the intelligence system. As more robots connect and more scenarios are covered, the system keeps learning and evolving, forming a real-world-driven intelligence flywheel.
Products

Litchi Lab · T1 Series Data Capture Kit

Pushing industry limits: extreme precision, ultra-low latency

Product Video

Omnimodal Perception

Product Demo

One Glove,
Control Multiple Dexterous Hands at Once

Millisecond-level latency reproduces every motion with precision, letting one glove coordinate teleoperation across multiple devices.

Product Specs

Technical Specifications

Precise to every metric, providing a reliable data foundation for Physical AI.

T1 Series Data Capture Kit

T1 Series Data Capture Glove

Lightweight, high-precision, and full-feature tiers covering the complete data pipeline from motion capture to tactile sensing.

Model Comparison
Category / Parameter T1 Lite T1 T1 Pro
Positioning Pure IMU lightweight motion capture Pure IMU high-precision motion capture IMU + tactile full-feature fusion
Key Advantage Lightweight, low cost, motion-capture focused Lightweight, high precision, motion-capture focused Complete data dimensions, ideal for high-quality robot training
Highlights High sampling rate · High pose resolution · Reliable end-to-end synchronization High sampling rate · High pose resolution · Reliable end-to-end synchronization High sampling rate · High pose resolution · High sensing resolution · High sensing density · Reliable end-to-end synchronization
Key Difference Hand pose, spatial motion, synchronized two-hand motion Hand pose, spatial motion, synchronized two-hand motion Hand pose + tactile pressure + contact and force events
Use Cases Teleoperation, gesture capture, motion imitation learning Teleoperation, high-precision gesture capture, motion imitation learning Dexterous manipulation training, embodied intelligence datasets, motion-force co-analysis
General Specifications

Sensor Configuration

IMU sensor configuration
T1 Lite: 7 IMUT1 / Pro: 11 IMU
IMU algorithm
Six-axis fusion algorithm
Sampling & output
IMU raw 1000 HzPose estimation 500 HzPro tactile 200 Hz (upgradeable to 400 Hz)
Motion capture method
Lite: 7/11 IMU pose captureT1 / Pro: 11 IMU pose capture
Static pose accuracy (Roll/Pitch)
≤ 0.05°

Wireless & Synchronization

Data transmission
Wireless
Transmission latency
2 ms
End-to-end sync error
< 1 ms
Wireless PHY rate
6.8 Mbps
Full wireless data rate
100 Hz
Left-right hand sync precision
Pro: microsecond-level
Magnetic interference resistance
Pro: supported (magnetometer-free reference / magnetometer downweighting)
Robot environment adaptability
Stable capture near motors and high-current wiring

Tactile Sensing (T1 Pro)

Tactile sensing
Full-palm flexible tactile array
Tactile capability
368 points standard (upgradeable to 600)
Point density
9 points/cm²
Minimum trigger force
30 g
Max single-point range
1500 g
Repeatability / consistency
±15% FS

Ego Capture Device

EgoDepth uses a hybrid architecture with 4 × wide-angle RGB surround cameras and 1 × structured-light stereo depth camera to help customers capture high-quality depth data.

Device Overview

Capture content
4-camera RGB + depth, IMU, audio
Typical use
High-precision operation at close range up to 1.5 m
Device weight
< 400 g
Architecture
4 × RGB wide-angle surround + 1 × structured-light stereo depth

Depth Module

Resolution & frame rate
1280×800@30fps848×480@30fps640×480@30fps
Field of view (FOV)
H 90° / V 65°
Spatial accuracy
< 0.5% @ 1m
Viewing angle
30° downward from horizontal

RGB Module (Single)

Resolution & frame rate
1920×1200@30fps (Max)1600×1200@30fps
Shutter type
Global shutter
Field of view (FOV)
H 165° / V 130°
Encoding
H.264 / H.265

IMU

IMU model
DCM630 9-axis
IMU sampling rate
500 Hz

Sync Accuracy

IMU-image sync
< 1 ms
External trigger
External trigger signal supported

SoC / Compute Platform

Processor
RK3576 8-core with 6T NPU
Memory & storage
2GB LPDDR4X + 64GB eMMC
USB ports
USB 3.0 Host + USB 3.0 Device
Wireless connectivity
Wi-Fi 6 + Bluetooth 5.4

Audio

Microphones
4 microphones
Speaker
1 W 8 Ω speaker

Storage & Development

Local storage
External TF card supported
SDK
Orbbec SDK
Software Specifications

Device Management

USB mode
UVC video + UAC audio + HID-channel IMU, with parameter set/get support
Wireless Wi-Fi mode
RTSP video/audio + proprietary TCP IMU, with network settings and parameter get/set
Device info retrieval
Read device SN, PID, device type, firmware version, etc.
Algorithm alignment calibration
YAML parameter file retrieval supported
IMU calibration parameters
IMU calibration YAML file retrieval supported
Firmware upgrade
Online firmware upgrade supported

Capture & Recording

Start/stop capture
One-click start and stop of data capture
Recording timer
Real-time recording duration display
Automatic session naming
Automatic naming and management of capture sessions
Dedicated recording directory
Each capture stored in its own directory
File formats
Video MP4, PTS CSV, IMU CSV, audio files

Video

Video stream capture
H.264 / MJPEG color streams, UVC protocol over USB, RTSP over network
Resolution/frame rate/format
Color resolution, frame rate, and format configurable
Image parameter control
AE, gain, exposure, white balance, and more settable/readable

Audio

USB audio capture
Audio streams captured via UAC in USB mode