Telexistence launches service that turns robot video into motion datasets

In short: Telexistence Inc. in Tokyo launched a service in January 2026 that turns robot camera footage into structured, searchable motion datasets.

What happened

Telexistence Inc. (TX Inc.), a robotics company based in Tokyo, launched a new service in January 2026. The service uses multi-joint robots and “deep learning” (a type of AI that learns patterns from lots of examples) to convert robot-recorded footage into motion datasets.

A motion dataset is like a well-organized library of movements. Instead of raw video, customers get data that is structured and searchable, so teams can quickly find specific actions or motion sequences. Telexistence says the output is tailored to each customer’s needs.

The company is aiming this service at robotics startups, large manufacturers, universities, and research labs in Japan, the U.S., and Europe. It offers a core package that generates datasets from robot-captured footage, and optional steps for data cleansing and curation, which means fixing errors and organizing the data to improve quality.

Pre-orders opened in September 2025. The pitch is that robotics teams need stable, customizable training data, but real-world robot video can be messy. For example, objects can be blocked from view (like a hand covering a camera), sensors can be noisy, and movements vary from one attempt to the next.

Why it matters

Better training data can make robots more reliable in everyday settings, like warehouses and factories. Think of it like teaching from a clear, labeled workbook instead of a pile of random photos and notes. If services like this reduce the time and cost of preparing data, more groups can build and test robot AI faster.

Source: TechCrunch AI