Person filming a video with a smartphone
Video Data for AI

Video Data Collection
for AI Training

Get short videos filmed by real people on their own phones in the countries you choose. Describe the action, set the pay and approve only the clips that fit.

Results in hours

Video tasks usually get first clips within minutes of going live.

200+ countries

Film short actions from everyday people in the countries you choose.

Review every clip

Approve only the videos that match your brief. Reject what does not.

200+

Countries

13,487

Tasks completed

195,000+

Registered workers

~4 Hours

Avg time to results

How It Works

How Video Data Collection Works

Launch a video task yourself and start receiving clips from real contributors, without long negotiations.

Person at a laptop writing video task instructions
01 · Create

Describe what people should film

Explain the action, camera angle, length and setting you need. Add an example video so workers see exactly what a good clip looks like.

  • Actions, activities or places
  • Example video for reference
  • Clear length and angle rules
Laptop with a world map on screen for country targeting
02 · Target

Choose countries and clip count

Select the countries your contributors should come from, set how many clips each person sends and how many submissions you need in total.

  • Any number of countries in one task
  • Several clips per submission
  • You set the total volume
Person recording a video of a scene with a smartphone
03 · Film

Contributors film on their own phones

Your task becomes available to workers in the selected countries. They record the scenes you described in their homes, streets and workplaces.

  • Different phones and cameras
  • Real homes, streets and workplaces
  • New people in every submission
Person watching video clips on a monitor
04 · Review

Approve only the clips that fit

Watch every submission, approve the videos that meet 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
Video Data Types

Types of Video Data You Can Collect

Set up a task for the exact clips your model needs. Contributors follow your instructions and film on their own phones.

Actions and Gestures

Movements performed by many different people

Contributors record gestures, hand signs, poses or simple actions you describe. Every person moves a little differently, which is exactly what your model needs.

  • Your own list of actions
  • Many people, many styles
  • Repeatable short clips
Create a video task
Person making a hand gesture toward a phone camera
First-Person Video

Point-of-view clips for physical AI

Workers film everyday tasks from their own point of view, with the phone held or mounted at eye or chest level. Used to train robotics and world models.

  • Egocentric view of real tasks
  • Natural hand movements
  • Homes and workplaces worldwide
Explore egocentric video dataCreate a video task
Hands cooking filmed from above in a first-person view
Everyday Activities

Real life, recorded as it happens

Collect clips of cooking, cleaning, shopping, exercising or working. Contributors film the activities you choose in their own environments.

  • Activities you define
  • Real homes and places
  • Local routines by country
Create a video task
Person cleaning or cooking at home for an activity clip
Hands and Objects

How people handle everyday things

Contributors film their hands opening, sorting, assembling or using objects. Useful for manipulation, product recognition and robotics models.

  • Close-up hand movements
  • Real objects, not props
  • Steps from start to finish
Create a video task
Hands assembling or sorting objects in close-up video
Talking to Camera

Short clips with face and voice

People record themselves speaking a phrase or answering a question on camera. One clip gives you video and audio for multimodal models.

  • Face and speech together
  • Scripted or free answers
  • Diverse ages and backgrounds
Create a video task
Person recording a selfie video speaking to the camera
Walkthroughs

Rooms, shops and streets on video

Contributors walk through apartments, stores, offices or streets while filming. Train models for navigation, mapping and scene understanding.

  • Indoor and outdoor routes
  • Local places by country
  • Steady, continuous shots
Create a video task
Person walking through a room or street filming with a phone
Task Example

Videos With Descriptions
in One Task

Ask contributors to add a short description of what happens in each clip. You get videos with context, ready for labeling or training.

Your setup

What you set up

Task settings

Choose Video collection, set how many clips each worker sends and write instructions with what to film, how long and from which angle.

  • Video collection task type
  • Number of clips per submission
  • Instructions with an example video

Your view

What you receive

Submission

Each submission arrives with the video clips and the worker's description, ready to review and approve.

  • Clips and text in one submission
  • Worker country on every submission
  • Approve or reject each set
Pricing

Estimate Your Video Data Collection Cost

You set the pay per task. Adjust the numbers to see what your project will cost, with the 15% platform fee included.

?
?
min
?
$
FAQ

Frequently Asked Questions

You set the pay for each task, starting from $0.10, plus a 15% platform fee. Longer or more complex clips need a higher pay rate. A campaign starts from $0.30, so you can test first.

Video tasks usually get first clips within minutes of going live, and the average time to results is about 4 hours.

You set the length, resolution, orientation and frame rate in your instructions, and add an example clip to show the result you expect. Short clips from a few seconds to a few minutes are easiest to film, upload and review.

Yes. Workers film everyday tasks from their own point of view, with the phone held or mounted at eye or chest level, so you see natural hand movements in real homes and workplaces. This data is used to train robotics and world models.

Yes. Ask contributors to describe what happens in each clip, and every submission will include the video and the text together.

Accept workers from all countries or pick continents and individual countries, and describe requirements such as age group, setting or phone type in your instructions. Each worker can complete your task only once, so every clip set comes from a different person.

State in your instructions that only the contributor may appear and that other people's faces, documents and screens must stay out of view. Watch each submission and reject any that breaks these rules.