One of the strangest new gigs in the AI economy does not involve a desk, a degree, or even a computer for most of the work. You strap a phone to your forehead, cook dinner, clean your kitchen, or organize a shelf, and a company pays you for the footage. These are egocentric data collection jobs, and they are hiring in dozens of countries right now. This guide explains what the work actually involves, what gear you need, the capture rules that decide whether you get paid, and the honest pay ranges by region.
What Do Egocentric Data Collectors Actually Do?
The job is simple to describe. You record everyday tasks from your own point of view: cooking a meal, washing dishes, folding laundry, wiping surfaces, packing boxes, stocking shelves, or doing basic warehouse and retail work. The camera sits at eye level so the footage shows your hands doing the task, exactly as a robot’s cameras would need to see it. If the concept is new to you, start with our explainer on what egocentric data is and why AI companies want it.
The workflow on most platforms looks like this. You pick tasks from a list in the platform’s app, record them following the capture rules, and upload the footage through the app the same day. A quality assurance team then reviews your clips, usually within 24 to 48 hours. Here is the part beginners miss: you are paid per approved hour of footage, not per hour you spent recording. Rejected clips earn nothing, which is why the rules in the next sections matter more than speed.
What Equipment Do You Need to Start?
The barrier to entry is low, which is exactly why these gigs spread so fast. Most platforms accept footage from a modern smartphone, provided it has an ultrawide camera (the 0.5x lens) so your hands and workspace stay in frame. Typical minimum devices named in job listings are the iPhone 12 or newer, Google Pixel 6 or newer, and Samsung Galaxy S21 or newer. The standard recording requirement is 1080p at 30 frames per second.
Beyond the phone, you need one accessory: a head strap or chest mount, which costs roughly $10-20 online. A head strap gives the truest eye-level view and is what most platforms prefer. That is genuinely the whole kit. No ring lights, no editing software, no PC. Before applying, check Claru AI’s first-person video collector listing for a concrete example of the device and footage requirements platforms ask for.
What Capture Rules Decide Whether Your Footage Gets Approved?
QA rejection is where new collectors lose most of their money, so learn these rules before your first session:
- Keep the camera at eye level, angled slightly down, so both hands stay fully visible in the frame for the entire task. Hands drifting out of frame is the most common rejection reason.
- Record complete, natural task sequences from start to finish. Clips often need to be at least 2 minutes long, and a task that stops halfway is usually worthless.
- No screen recordings and no filming a screen. The footage must be of the real physical task.
- No fake or purposeless motion. Waving your hands around, repeating a gesture with nothing in them, or staging empty activity gets flagged and rejected.
- No third-person filming. Having someone else hold the camera and film you defeats the entire purpose and is treated as a rule violation.
- Fraudulent footage is not paid, period. Reused clips, other people’s videos, and AI-generated submissions can get your account banned, not just rejected.
A practical tip that separates good collectors from frustrated ones: do a 30-second test clip, watch it back, and check that your hands are visible the whole time before recording a full batch. Five minutes of checking saves hours of unpaid re-recording.
How Do You Get Started, Step by Step?
- Step 1: Pick a platform that operates in your country and matches your rate expectations. Our comparison of the best platforms that pay you to train AI groups the egocentric specialists separately from general AI training sites.
- Step 2: Apply and complete onboarding, which usually includes an identity check. Legitimate platforms verify who you are because they are buying data with usage rights attached.
- Step 3: Pass the test recording. Most platforms make you submit a short trial task before giving you paid work, and this trial is often unpaid.
- Step 4: Record in batches. Pick a few tasks, set up once, and record them back to back so your setup time is spread across more footage.
- Step 5: Upload the same day through the app, while you still remember which clips might have problems, and check your QA results before recording more of the same task.
How Much Do Egocentric Data Collection Jobs Really Pay?
Pay varies enormously by country, and honest ranges matter more than the headline numbers in ads. Claru AI lists first-person video collectors at $15-30 per hour for the US and Canada. At the other end, egocentric gigs posted on Upwork for projects in Latin America and Asia typically run around $5-6 per hour, and Toloka’s first-person video trainer roles pay up to about $5 per hour worldwide. This breakdown of getting paid to record household chores for AI data platforms shows how widely the same work is priced in different regions.
Remember that these are rates per approved hour of footage. If a third of your clips get rejected while you are learning, your effective rate drops by a third. Experienced collectors protect their rate by mastering the capture rules first and treating QA feedback as free training. Also weigh the rate against your local alternatives: $5 an hour is poor pay in Berlin and genuinely competitive pay in many other places, which is precisely why this work is distributed the way it is.
What Are the Honest Downsides?
This is gig work in its purest form, and it comes with the standard fine print. Projects are campaign-based: a platform may need thousands of hours of kitchen footage this month and none next month. There are no guaranteed hours, no contract, and no benefits. Some platforms cap how much one contributor can record in a single environment, for example around 100 hours, after which you need a new location or a different task category to keep earning. Unpaid assessments and test tasks are common across AI training work, and egocentric collection is no exception.
There is also the bigger question of what the footage trains. This data teaches robots and AI systems to perform physical tasks, and some collectors are uneasy about that. Only you can decide where you stand, but go in with your eyes open. As one stream among several, though, it is a legitimate way to earn from skills you already have, much like the other routes in our guide to making money as a small creator: none of them is a full career on its own, and together they add up.
Frequently Asked Questions
Do I need experience or qualifications to get egocentric data jobs?
No degree or work experience is required. Platforms care about three things: a compatible phone with an ultrawide camera, the ability to follow capture rules precisely, and passing their ID check and test recording. Careful, rule-following beginners routinely outperform people who rush and get half their footage rejected.
Why do platforms pay per approved hour instead of per hour worked?
Because they are buying a dataset, not your time. Footage with hands out of frame, incomplete tasks, or rule violations has no training value, so the business model only pays for usable material. It feels harsh at first, but once your approval rate climbs above 90 percent, the distinction mostly stops mattering.
Can I do this work from any country?
Not every platform hires everywhere, and rates are set by region. US and Canadian collectors get the highest listed rates on specialist platforms like Claru AI, while global platforms such as Toloka accept contributors in many more countries at lower rates. Check each platform’s country list before investing in a mount or doing an unpaid test.
Is it safe to record inside my own home for these platforms?
It can be, if you follow the consent and privacy rules: tell household members, keep children and bystanders out of frame where required, avoid documents and screens with personal information, and read the data usage terms before you start. If a platform’s terms are vague about how footage can be reused or resold, treat that as a warning sign.
Last updated: October 2026.
