Back to feed
News Story
APriority78
机器之心
1 sources

HOMIE Gen2 Released! Unlocking the Scaling Law of Human Experience

Ropedia has released the new multimodal capture system HOMIE Gen2, designed to scale human experience into training data for robots. The company proposes a 'Human Experience Scaling Law,' suggesting that robot capabilities will improve with the scalable growth of high-quality human experience. HOMIE Gen2 features immersive capture capabilities and synchronized multi-modal recording, providing training-grade data for embodied AI.

SynthePulse Insight · AI deep reading

HOMIE Gen2: Using a Headset to Collect 'Real-World Data' for Physical AI

Version 1 · 1 source

Singapore-based company Ropedia releases its fourth-generation head-mounted data collection system, HOMIE Gen2, aiming to address the shortage of training data for physical AI with human first-person 'experience data.'

  • Ropedia launches HOMIE Gen2, a 380-gram head-mounted data collection device with 360-degree visual coverage, 1080p/30fps video, and four-channel spatial audio.
  • The device processes over 10 data types through its 'Human Experience Engine,' with the company claiming up to 96% accuracy in capturing and interpreting human activities.
  • HOMIE Gen2 is the fourth-generation hardware developed in 12 months, following a pre-Series A funding round that brought Ropedia's total funding to $30 million.
  • The company believes physical AI is at a stage similar to before the large-scale deployment of autonomous driving, and human data can bridge the gap caused by insufficient robot deployment.
Open section navigationPhysical AI's Data Gap and Ropedia's Answer

Physical AI's Data Gap and Ropedia's Answer

Ropedia released HOMIE Gen2 on August 26, 2026, a head-mounted data collection system designed to record real-world human activities and convert them into training data for robots and physical AI models. In the announcement, CEO Zhaoxi Chen stated that large language models have proven that scaling training data can unlock new levels of intelligence, and physical AI follows a similar scaling principle, but its training data must capture the complexity of the real world.

Chen noted that traditional fixed camera setups and controlled demonstrations can capture individual actions but struggle to provide the diversity and in-the-wild data needed to train systems for the physical world. HOMIE Gen2 is designed to enable data collection in natural environments rather than recreating them in dedicated data collection facilities.

Hardware Specifications and Capture Capabilities

The HOMIE Gen2 weighs 380 grams, offers 360-degree visual coverage, records 1080p video at 30 frames per second with a global shutter (capturing the entire image at once rather than scanning line by line), and features four-channel spatial audio. The device also includes versatile ports that support synchronization of multiple devices, allowing simultaneous recording of the environment from different people and perspectives.

The system is self-contained and hands-free, suitable for wearing in settings such as homes, factories, and retail. Ropedia states that multiple HOMIE Gen2 units can be used simultaneously to record the same environment from different viewpoints.

Data Processing and the 'Human Experience Engine'

Recorded content is processed through Ropedia's 'Human Experience Engine,' which synchronizes and structures over 10 types of information, including video, audio, depth, motion, body pose, and spatial data. The company claims tests show the system achieves up to 96% accuracy in capturing and interpreting real-world human activities.

HOMIE Gen2 builds on Ropedia's Xperience-10M dataset, released in March, which contains over 10 million interaction segments derived from 10,000 hours of first-person recordings.

Funding Background and Industry Positioning

This release follows Ropedia's pre-Series A funding round, bringing the company's total funding to $30 million. HOMIE Gen2 is the fourth-generation capture hardware developed by the company in the past 12 months.

Chen compared the current state of physical AI to the stage before large-scale deployment of autonomous driving: 'Autonomous driving has millions of cars on the road generating real-world data, but physical AI does not yet have robot deployments at that scale. Human data helps bridge this gap.'

Credibility boundary

The information in this article is primarily sourced from Ropedia's official announcement, as republished by The AI Insider. All data regarding product performance (such as 96% accuracy) and dataset scale (such as 10 million interaction segments) are company claims and have not been independently verified.

Insight takeaway

HOMIE Gen2 represents a new approach to data collection for physical AI: using human first-person 'experience' to fill the data gap caused by insufficient robot deployment. However, its actual effectiveness and the 96% accuracy claim still require independent verification.

Sources for this version

  1. Ropedia Launches HOMIE Gen2 to Scale Real-World Data for Physical AIThe AI Insider

Primary report

机器之心

Primary source