OriginFlow | Physical AI Multimodal Interaction Data Infrastructure

PHYSICAL AI / DATA

ORIGINKITGEN 1.0 / HUMAN INTERACTION CAPTURE

Turn real-world actions

Into machine-learnable data

OriginKitGen 1.0 combines three devices—Glass, Band, and Space—to synchronously capture first-person video, human motion, and 16-channel surface EMG during real-world work, giving robotics and Physical AI teams traceable multimodal interaction data.

Learn About OriginKitGen 1.0

Contact the Team

Products & Services

OriginKitGen 1.0

Data Capture Services

01 / ORIGINKITGEN 1.0

Multimodal Data Capture

First-person video, head motion, surface EMG, and wrist motion are captured in sync, with device control, local recording, data storage, and operation feedback handled in one workflow.

01 / GLASS

First-person Vision and Head Motion

Records real-world work from the operator’s point of view, creating a visual context that connects actions with their environment.

02 / BAND

Surface EMG and Wrist Motion

Captures neuromuscular activity as actions happen, adding signals that vision cannot observe directly.

03 / SPACE

Unified Control and Local Recording

Manages device status, capture workflows, and local data in one place, so every task can be reviewed and traced.

02 / WEAR TO CAPTURE

Capture Real-World Tasks and Workflows

Device Setup

Configure and wear Glass, Band, and Space for the task.

Quick Calibration

Confirm device connections, timeline alignment, and capture status.

Start Capture

Record actions, signals, and environmental changes throughout the real-world task.

Status Feedback

Use Space to view device status and local feedback.

Data Preview

Review and verify the completed task on a unified timeline.

03 / CAPTURE & DATA

Multimodal data across video, semantics, pose, and force

Neuromuscular signal capture, multimodal time synchronization, and interaction-state reconstruction support the full workflow from field recording to downstream analysis.

High-Density Neuromuscular Signal Capture

Use high-density surface EMG to complement neuromuscular activity that cannot be observed directly through vision.

Multimodal Time Synchronization

Align first-person video, motion data, and surface EMG on a unified timeline for correlation and review.

Interaction-State Reconstruction

Under defined algorithms, calibration, and task conditions, estimate pose, motion intent, force, and contact state.

04 / ORIGINFLOW

Starting from Human Experience

Learning Real Actions,

Extending Human Capability.

Starting from Human Experience

Learning Real Actions,

Extending Human Capability.

Founded in 2025, OriginFlow focuses on building multimodal human interaction capture systems and real-world data infrastructure for Physical AI.

HUMAN EXPERIENCE

PHYSICAL AI

REAL-WORLD DATA

05 / CONTACT

Learn About the Product—Contact Us

If you want to explore OriginKitGen 1.0, schedule a product demo, or discuss real-world capture partnerships, we’d love to hear from you.

contact@originflow.ai

Zhongguancun, Haidian District, Beijing · Yingu Building, Room 802

Multimodal Human Interaction Capture Systems · Real-World Data Infrastructure