Egocentric and robotics video
First-person (egocentric) video, workplace task footage, robot teleoperation recordings and sensor-synchronized video.
- Measured in
- Hours of footage, task categories, number of distinct operators and environments.
- Who buys
- Robotics companies, humanoid and manipulation model builders, physical AI labs.
- Where it comes from
- Manufacturers, kitchens, warehouses, trades and training providers, research labs and wearable camera programs.
- Rights to clear
- Employee consent and notices, union agreements, client site permissions and safety footage policies.
- Privacy
- Faces of workers and bystanders, screens and documents in frame, customer premises.
- Pricing signal
- Priced deal by deal. Labeled, multi-environment footage commands the most.
- Robotics and physical AI teams need first-person video of people doing real tasks with their hands.
- Footage from fixed workplace cameras, wearables and teleoperation sessions is rarely available online, which makes it valuable.
- Task labels, such as which step of a procedure is happening, raise the value sharply.
- Workplace footage needs worker consent and removal of anything that identifies customers or bystanders.
Why this data is scarce
Language models could learn from the web. Robots cannot: there is very little online video showing, from the doer's point of view, how a skilled person assembles, cooks, repairs or packs something. Teams building manipulation and humanoid models are therefore buying footage directly from businesses where skilled hand work happens every day.
What buyers value
- First-person or close over-the-shoulder views of hands and tools
- Many operators and many environments, not one person at one bench
- Task and step labels, including failures and corrections
- Synchronized sensor data where available (depth, IMU, force)
- Consistent capture settings
Consent and workplace issues
Recording workers for a third party's model training is different from recording them for safety or quality. Workers should be told what is being collected and why, and many deals include payment or opt-in. Union agreements may require bargaining. Faces, name badges, screens and customer premises usually need to be excluded or blurred.
Starting small
Many buyers fund a pilot: a few dozen hours across defined tasks, to test whether the footage improves their models before a larger collection. That is often the best route for a business with skilled operators but no existing archive.
Questions
We don't have footage yet. Can we still sell?
Yes. Some buyers fund new collection at your site under a defined protocol. That is a services agreement as well as a data license.
Do workers have to agree?
Buyers expect informed consent from anyone recorded, and many programs pay participants. Check union and employment obligations first.
Is CCTV footage useful?
Usually less so. Fixed overhead cameras rarely show hand detail, and they raise more privacy issues. Close task footage is what buyers want.
Related guides
Related data types
Sources
Need egocentric and robotics video, or hold it?
Tell us what you need or what you hold. We check rights and privacy first, then introduce vetted counterparties. You license directly with each other.
- We reply to every enquiry within two working days.
- No upfront fees for sellers
- We never hold your raw data