Egocentric Video 路 Physical AI200+ countries 路 First-person tasks

First person.Real hands.Physical AI.

Get first-person videos of real hands-on tasks filmed by people in their own homes and workplaces, in the countries you choose. Built for robotics, physical AI and world models.

200+

Countries

13,487

Tasks completed

195,000+

Registered workers

~4 Hours

Avg time to results

How It Works

How Egocentric Video Collection Works

Launch a first-person video task yourself and start receiving footage of real tasks, without long negotiations.

Person at a laptop writing egocentric video filming instructions
01 路 Create

Describe the task to film

List the activity, the steps, the camera position and the length you need. Add an example clip so workers see the exact point of view you expect.

  • Activity and steps to perform
  • Camera position and length
  • Example clip for reference
Laptop with a world map on screen for country targeting
02 路 Target

Choose countries and volume

Select the countries your contributors should come from and how many clips you need. Different homes, tools and habits make your model work beyond the lab.

  • Any number of countries in one task
  • Different homes and workplaces
  • You set the total volume
Person with a phone on a chest mount filming a kitchen task hands-free
03 路 Film

Contributors film from their own point of view

Workers mount their phone on their chest or head, keep both hands free and record the task from start to finish in their own environment.

  • Phone on a chest or head mount
  • Both hands visible and free
  • Full task from start to finish
Person watching first-person video clips on a computer monitor
04 路 Review

Approve only the clips that fit

Watch every submission, approve the footage that meets your requirements and reject the rest. Keep your dataset clean from the start.

  • Watch every clip
  • Approve or reject each submission
  • Clean data for training
Egocentric Data Types

Types of Egocentric Video You Can Collect

Set up a task for the exact activities your robot or model needs to learn. Contributors film them from their own point of view.

Household Chores

Cleaning, laundry and tidying up

Contributors film themselves wiping tables, folding clothes, loading dishwashers and putting things away. Teach home robots how real chores are done.

  • Everyday chores you choose
  • Real homes, not studios
  • Full task from start to finish
Create an egocentric video task
First-person view of hands folding laundry on a table
Kitchen and Cooking

Real cooking from the cook's point of view

Workers record preparing simple meals, washing produce, opening packages and using kitchen tools. Rich hand movements with many different objects.

  • Simple, safe recipes
  • Many tools and ingredients
  • Local kitchens by country
Create an egocentric video task
First-person view of hands cutting vegetables in a kitchen
Hand-Object Interaction

How hands grasp, move and use things

Contributors pick up, open, turn, pour and place everyday objects as you describe. Close-up footage for grasping and manipulation models.

  • Your own list of actions
  • Close view of both hands
  • Many object shapes and sizes
Create an egocentric video task
Close first-person view of hands opening a jar
Tool Use

Simple repairs and assembly, step by step

Workers film assembling furniture, tightening screws, changing batteries or fixing small items. Show models how people use tools in real life.

  • Safe, simple tools only
  • Step-by-step actions
  • Real objects and parts
Create an egocentric video task
First-person view of hands using a screwdriver for a simple repair
Packing and Sorting

Sorting, packing and organizing items

Contributors sort laundry, pack boxes, organize shelves or group items by type. Useful for logistics, retail and warehouse robotics.

  • Sorting rules you define
  • Boxes, bags and shelves
  • Repeatable short sessions
Create an egocentric video task
First-person view of hands packing items into a box
Indoor Navigation

Moving through real homes and rooms

Workers walk through their home, open doors, climb stairs and move between rooms while filming. Train models to understand indoor spaces.

  • Rooms, doors and stairs
  • Steady walking pace
  • Different home layouts by country
Create an egocentric video task
First-person view of a hallway and open door inside a home
Task Example

Clips With Step Descriptions
in One Task

Ask contributors to list the steps they performed with approximate times. You get footage with action labels, ready for annotation or training.

Your setup

What you set up

Task settings

Choose Video collection, describe the activity, camera position and length, and ask workers to list each step with its start time.

  • Video collection task type
  • Activity and camera rules
  • Step list with times

Your view

What you receive

Submission

Each submission arrives with the first-person video and a list of steps with approximate times, ready to review and approve.

  • Video and step list in one submission
  • Worker country on every submission
  • Approve or reject each clip
FAQ

Frequently Asked Questions

Workers attach their phone to a chest or head mount so both hands stay free, then record the task from start to finish in their own home or workplace. You can require a mount in your instructions.

Household chores, kitchen and cooking, hand-object interaction, tool use and simple repairs, packing and sorting, and indoor navigation through rooms, doors and stairs. You list the activity, the steps and the camera position.

You set the pay for each task, starting from $0.10, plus a 15% platform fee. Longer filming and mount requirements need a higher pay rate so workers take the task. A campaign starts from $0.30.

You decide in your instructions. Clips of one to ten minutes are easiest to film, upload and review; split longer sessions into several clips.

Yes. Ask contributors to list each step they performed with an approximate start time, and every submission arrives with the video and the step list together.

Every task states that the videos will train AI, and only adults who choose to take part film. No other people, children, documents, screens or addresses may appear, and clips that break these rules are rejected.

Accept workers from all countries or pick continents and individual countries. Add requirements such as owning a phone mount, a type of home or a specific appliance; each worker completes your task once, so every clip comes from a different person and home.