Ropedia unveils HOMIE Gen2, a wearable capture device built to record human "experience" data for training robots.
The AMW Read
Incremental hardware/product update from an already-tracked Physical-AI data-infrastructure player, relevant to the embodied-foundation-model data layer but sub-segment in impact.
Ropedia unveils HOMIE Gen2, a wearable capture device built to record human "experience" data for training robots.
Ropedia, a Singapore-based Physical AI company co-founded by CEO Zhaoxi Chen and CTO Fangzhou Hong, has launched HOMIE Gen2, a roughly 380-gram wearable multimodal data-capture device. The unit combines four panoramic cameras for 360-degree coverage, four-channel spatial audio, and multi-sensor synchronization to within 50 microseconds, and can record continuously for about 13 hours without external setup. Ropedia says Gen2 deploys roughly 10 times faster and cuts collection cost to about one-twelfth of its prior hardware generation. Chief scientist Ziwei Liu, an NTU associate professor, frames the effort around what the company calls a Human Experience Scaling Law: robot capability should scale with the volume of high-quality, multimodal human experience data, not video alone. Ropedia previously released Xperience-10M, a roughly 1-petabyte dataset of 10 million real-world interaction clips and 10,000 hours of audio-paired first-person video, and says it has supplied data to more than 20 robotics and foundation-model teams globally.
The launch targets a well-documented gap in embodied AI: models trained on internet-scale text and video can describe a task fluently but often fail at physical execution, since action, timing, and environmental feedback are rarely captured together at the fidelity robots need. As foundation-model-style training extends into physical tasks, competition among data-infrastructure vendors is shifting from whether raw motion data can be collected to whether captured sessions align action, intent, environment, and multimodal signal completely enough to be training-ready without heavy post-processing.
For builders and investors, wearable-capture infrastructure is emerging as a distinct layer beneath embodied foundation models, alongside teleoperation and simulation data pipelines. Per the AI Market Watch index, Ropedia is tracked with $30M in total funding since its 2025 founding (the index covers roughly 5,000 companies, not a census); a maturing hardware line and an expanding customer base of robotics labs are signals worth watching for follow-on fundraising or dataset licensing deals in this sub-segment.
