Mount your phone on a chest or head strap and film yourself doing real household work from your own point of view, at your natural speed, talking through what you are doing. The most important rule: keep both hands in frame while you work.
- Role — Video capture contributor — records and narrates first-person footage of their own household activities
- Location — Remote, in the contributor’s own home
- Deliverable — Approved video clips, 5 minutes to 2 hours each, one continuous take, narrated
- Equipment — Approved smartphone (iPhone 11+, Pixel 6+, Galaxy S21+) and a chest, neck, or head strap
- Experience — None required — no degree and no prior gig work. Onboarding and one unpaid practice clip provided.
What this work is
We are building AI that understands the physical world the way people do — how we fix things, build things, care for a home and move through it. To do that we need real footage of everyday life shot from a first-person point of view: video from a camera worn on the body,
showing the world through the contributor’s own eyes and hands.
Every clip is narrated. Contributors describe what they are doing out loud while they do it. The spoken track is not a commentary added on top — it is half of what we are collecting. A model learns far more from footage of a repair paired with the person saying which part they are removing and why than it does from the same footage in silence. A clip with no narration is not usable, however good the picture is.
Contributors create the original footage and audio themselves. This is data collection, not labeling or annotation — they produce the raw material rather than review anyone else’s work.
This project does not collect kitchen or cooking work, and does not collect laundry. Those are the two most common activities in existing footage libraries, so they are deliberately out of scope here. The work we need is repair, assembly, yard, vehicle, technical, and care tasks.