Headset on, a controller in each hand. Somewhere across the country a robot arm copies every move you make as it lifts a mug onto a shelf. You do it again, from a slightly different angle, and again.
It sounds strange, and it is real work. Companies training robots need hours of recordings of people doing ordinary physical things, and they hire people like you to produce them.
Why should you care now? Because ordinary jobs feel shakier every year. Warehouse shifts get automated. Office tasks go to software. Your rent rises while the hours offered keep shrinking. Here is a chance to be paid by the very industry building that future, while it still needs human hands.
The demand is early, and early work goes to the people already approved, with clean records, when the next project opens. Applicants who start later queue behind them.
Kit costs you nothing to begin, or about $250 if you buy your own capture gear. Expect roughly $800 a month at the start, and first money tends to land two to six weeks in.
Read the paperwork with care. A posting's data clause can decide whether you get paid once for a motion or keep any rights to it, and on this work it deserves more attention than a $12 an hour job would normally get.
Today, open one teleoperation posting and read its data clause before you read its pay rate.
Why take it seriously as an income path? Because the rates are published on public job boards, so you can check them before you give up a single evening. Why be careful? Because this is contract work, and you will never own it as a business. You are selling your hours, someone else owns the asset those hours build, and the project ends when the dataset is full. Everything below keeps that distinction in view, because it changes what the work is good for in your life.
Language models had the internet. Robots have nothing like it. Nobody has a vast archive of a hand turning a doorknob, recorded from the right angle with synchronised force and position data. It has to be made, one demonstration at a time, by people.
The industry says this openly. The Robot Report's August 2026 survey of physical AI infrastructure platforms describes one major provider's approach as combining "centralized data factories, distributed human collectors, real robot systems, and robotless egocentric collection, with multimodal annotation and internal policy fine-tuning". Read that phrase slowly, because a job description is hiding inside the industry vocabulary. "Distributed human collectors" means people at home with a phone, people like you. "Robotless egocentric collection" means recording a task from your own point of view with no robot in the room. "Data factories" means shift work in a building full of robot arms.
Those three phrases map almost exactly onto the three tiers of work open to you, and the pay gap between them is roughly fivefold.
Which of those three could you realistically do from where you live today? Hold that answer, because it decides most of what this work can pay you.
The shape of this job comes straight out of how the last fifteen years of machine learning went, and knowing that history helps you guess where it goes next.
Image recognition was solved, in commercial terms, by building a very large labelled dataset and then scaling models against it. Language followed the same pattern with a huge head start: the dataset already existed, because people had spent thirty years typing the internet. Nobody had to commission it. The scaling recipe worked because the data was free and plentiful, and a whole generation of the field took away one lesson: capability follows data volume.
Robotics inherited the recipe and none of the data. There is no web-scale collection of physical manipulation. What existed instead was thousands of small academic datasets, each recorded on a different robot, in a different lab, with different cameras and different conventions, and they could not be pooled. A model trained on one lab's robot did not carry over to another's.
Two things changed around the middle of the decade. Shared data formats and open tooling made pooling across robot types practical, so a demonstration recorded on one arm became useful to someone with a different arm. And the field showed that recordings of people, captured from a first-person view, carry enough signal to be worth training on even with no robot involved. That second finding created the remote tier of this work. It turned robot data collection from something only a lab could do into something you could do with a phone and a specification.
So the bottleneck moved from hardware to logistics. Robots are no longer the scarce resource. Organised human hours are: people producing consistent recordings across enough environments and enough variation. That is a staffing problem, and it is why these listings exist for you to apply to.
So the demand is real and somebody is paying for it. What that means for you comes down to which tier you land in.
Every figure in this section comes from a live listing on a company's own job board. None come from a salary aggregator's estimate, which matters because the aggregate figures for this role mix together very different jobs.
Two facts fall out of that table. Pay follows the hardware you are trusted with, and your experience barely moves it. And the entry tier pays poorly, which hardly anyone advertising this work says clearly, so let me say it to you: at $10 an hour, phone capture is worth doing to build a track record with a platform, or as flexible fill-in income, and it is a poor reason to reorganise your life.
Even the $10 an hour phone capture tier has a use when money is tight. A few evenings folding towels on camera to a spec can build the buffer that pays your next electricity bill in full, and that small record gets you considered for the better tiers. If approved captures add up to 800 a month, count only work that has actually been paid when you decide what to set aside.
If you earned $10 an hour for a month of evenings, what would that money be for, specifically? A clear answer tells you whether the entry tier is worth your time or just your tiredness.
Even the entry tier can matter at home. A month of evening captures could cover a child's after school club or the dentist bill you have been putting off. Climb a tier and the same hours start funding a holiday instead of a crisis.
Headline rates attached to famous companies spread quickly in this field. A widely repeated figure puts teleoperation work at one large automaker at $48 an hour, and another claim has a data platform paying $50 an hour for people to record themselves. Both appear in secondary write-ups and social posts. I could not verify either on the hiring company's own careers page, so treat them the way you would treat any income claim with no primary source behind it: possible, unconfirmed, and no basis for your plans.
Salary aggregators report an average around $28 an hour for "robot teleoperation" in the United States as of mid-August 2026, with most of the range between roughly $22 and the mid thirties. That number points the same way as the on-site listings above, but averages for a job title this new lump together truly different roles, so the specific listing in front of you is better evidence than any average.
The practical rule for you: apply to the posting, read the stated range, and ignore the number you saw on social media.
A data buyer is collecting a dataset with consistent properties, and the chore is almost incidental. Those properties are things like a fixed camera position, a stated frame rate, particular lighting, a defined start and end state for the task, a required number of repetitions, no faces or identifying documents in frame, no other people, and sometimes a specified hand to use. A recording that captures a beautiful, natural, useful-looking demonstration and breaks one of those rules is worth nothing, because it cannot be pooled with the rest.
That is why the first week is where most people quietly drop out. The task looks trivial, the spec looks like bureaucracy, and then your submissions get rejected and your real hourly rate falls well below the posted one. The operators who earn the posted rate treat the spec as the deliverable and the chore as a detail.
Three habits carry most of the benefit for you. Read the spec end to end before your first recording, then record one clip and submit it alone for feedback before batching twenty. Set up a fixed physical rig, a mount position, a marked spot on the floor, the same lamp, so compliance comes from your setup and you never have to remember it. And keep a short log of what was rejected and why, because rejection reasons repeat, and you can avoid the second one.
When you last followed a long set of instructions, did you read them all first, or start and check back when stuck? This work rewards the first habit and quietly punishes the second.
Listings use a small vocabulary for the kinds of data they want, and once you recognise the words you can tell what a role involves before you apply. A posting from one humanoid robotics company for a Data Collection Operator describes operating humanoid robots "during data collection activities, including locomotion, manipulation, teleoperation, and human-robot interaction scenarios". Those four terms cover most of the field.
Manipulation is the bulk of the work and the most valuable per hour. Grasping, placing, opening, pouring, folding, inserting. It is hard because contact is hard: the moment a gripper touches an object is where the useful signal lives, and also where cameras get blocked and force feedback matters most.
Locomotion is walking, balancing, stairs, uneven ground, recovering from a shove. Mostly limited to on-site work with legged hardware, and rarely something you would contribute to as a remote collector.
Human-robot interaction is handing an object to a person, working alongside someone, responding to being interrupted. It is growing quickly, because a robot that only works in an empty room cannot be deployed. It needs at least two people present, which makes it expensive to organise and therefore better paid.
Long-horizon and multi-step tasks are the current frontier: clear the table, then load the dishwasher, then wipe the surface. Each step is easy and the whole sequence is hard, because the model has to keep track of a goal across minutes where it is used to seconds. Expect these specs to be the fussiest and the most likely to change mid-project.
Running across all four is what buyers call coverage, and it is what most collectors misunderstand. A dataset of a thousand flawless mug pickups in one kitchen is close to worthless. Variation is the product: different lighting, different surfaces, different object positions, cluttered scenes, unfamiliar objects, your left hand as well as your right. When a spec asks you to move a lamp between takes or record the same task at three times of day, that instruction is the entire point of your work.
What follows feels backwards. Failure is often requested. Specs increasingly ask for the fumble and the correction, the grasp that slips and the recovery, because a model trained only on success has no idea what to do when something goes wrong. New collectors instinctively delete those takes. Read the spec before you do.
How many different rooms, surfaces and lighting setups could you offer from your own home? That variety is quietly part of what you are selling.
The Kit for Remote Capture
That settles the money side. What follows is what has to sit on your desk before anyone can send you a shift.
The on-site tier supplies its own hardware. For remote work your equipment needs are small but specific, and getting them wrong is the most common reason batches get rejected.
A hands-free head or chest mount. The listings ask for "hands-free, first-person" video, which rules out holding the phone. A chest harness is steadier and cheaper; a head mount gives a viewpoint closer to where a robot's cameras sit and is usually what specs want. What matters most is repeatability: the same mount at the same angle every session, so your captures pool with your own earlier work.
A phone that records long clips without overheating. Long high-frame-rate capture heats phones, and a phone that drops frames or stops recording at minute eleven will fail a spec silently. Test a full-length capture before your first paid batch.
Storage and upload bandwidth. This is the cost nobody budgets for. High-frame-rate first-person video is large, batches run to many gigabytes, and buyers want your original file with no re-encoding. If your upload speed is slow, that time is unpaid, and it can quietly halve your real hourly rate at the entry tier.
Controlled, boring lighting. Consistent artificial light beats good natural light, because daylight changes between takes and the spec cares about consistency far more than beauty. A cheap lamp in a fixed position solves more rejections than any camera upgrade.
A marked setup. Tape on the floor for where you stand, tape on the counter for where objects start. This makes compliance a feature of your kitchen, so you never have to concentrate on it.
Where a role uses an instrumented rig, such as the UMI-style gripper in the Argentina listing, the buyer ships it to you. Read the terms on that hardware carefully: you are usually responsible for it, and returning it at the end of the contract is a condition of your final payment.
How fast is your home upload, honestly, on a weekday evening? Test it before you apply, because at $10 an hour slow uploads come straight out of your pay.
Who Is Actually Hiring
You can capture the work now. The next question is who you point that setup at.
The buyers fall into three groups, and it helps you to know which one you are talking to.
Data platforms and collection companies. These are businesses whose whole product is datasets and annotation services for robotics teams. They run the labs, recruit the distributed collectors, and hold the contracts with model developers. Most of the listings you find will come from here, and OpenTrain AI is one example. Their work is steady, their specs are strict, and they are the most likely to have something for you as a beginner.
Robot developers hiring directly. Humanoid and manipulation companies post their own data collection and teleoperation roles, sometimes titled Data Collection Operator or Robot Operator. These tend to be on-site, better paid and more competitive, and they often expect you to be at ease around expensive hardware.
Annotation and infrastructure vendors. Companies that provide the tooling also staff human work around it. Roles here lean toward reviewing, labelling and quality-checking other people's captures, with less recording of your own, which is a different and often more sustainable job.
To find real listings, go to company job boards before aggregators, because aggregators lag and duplicate. Searching the specific vocabulary works better than searching "robot jobs": teleoperator, data collection operator, egocentric video, demonstration data, physical AI, UMI gripper.
Spotting the Scams, Which Are Already Here
Any income category with a real published rate and a low skill floor attracts fraud within months, and this one has reached that point. Please read this section twice.
The reliable signals are old ones. Nobody legitimate asks you to pay for training, certification, equipment access or a "collector kit" before you can start earning. Nobody legitimate needs your bank details before an offer exists. A real contract states a rate, and a posting that describes your earnings only as a possible monthly total is avoiding the hourly number for a reason.
Two signals are specific to this field. First, real collection work comes with a written specification, and a long one; an offer with no spec attached is no data job at all. Second, legitimate work has narrow geographic eligibility, because of tax status, data protection law and shipping. A listing that will hire anyone anywhere with no eligibility conditions is either fake or paying less than it says.
One more, less obvious: be wary of anything asking you to record inside a workplace, a school, a hospital or anywhere with other people who have not consented. That request shows the buyer is careless, and the consequences of that carelessness land on you. They walk away clean.
If a "recruiter" asked you for a kit fee tomorrow, would you have a rule ready, or would you weigh it up in the moment? Decide your rule now: no payment from you, ever, before paid work exists.
Getting Through the Application
The hiring process is lighter than the pay range suggests, which is the main reason this work is worth an application even if you are sceptical.
Screening for the remote tiers usually means a short form, an eligibility check on your location, and a sample capture. The sample is the real filter. You are sent a specification and asked to produce one compliant clip, and the reviewer cares little about how well you fold a towel. They are checking whether you can follow a written spec exactly, because that predicts whether you will be cheap or expensive to work with.
Treat your sample that way. Follow the spec literally, including the parts that seem pointless. If something is unclear, ask before recording, and say how you have interpreted it. A candidate who asks one precise clarifying question looks low-risk; a candidate who submits a confidently wrong interpretation looks like someone whose batches will need checking twice.
For the on-site tier, expect the extra questions to be practical: shift availability, whether you can commit to the stated minimum hours, whether you can stand and move through a facility for a full shift, and how you handle repetitive work. Nobody is testing your robotics knowledge. They are testing whether you will still be there in week six, because turnover is their expensive problem.
Two things help your application far more than you would expect. Any precision work in your history, whether that is machining, lab work, surgery, music, sewing or competitive gaming, is worth naming. And being willing to take an unpopular shift is real leverage at a facility that has to staff a midnight crew.
If you are willing to take the midnight crew at a lab paying $30 to $55 an hour, you hold leverage most applicants never use. If you land the work and the hours fit your household, a few of those shifts a month could mean your daughter's football boots and the class trip are sorted before payday. Name the precision work in your history on the form and make your shift availability clear.
What in your own past counts as precision work, even if you never thought of it that way? Sewing, music and gaming all count, so write yours down before you fill in the form.
How It Compares to the Alternatives
It helps to set this against the other things you might do with the same hours, because the honest answer depends on which tier you can reach.
Against delivery and rideshare work, on-site teleoperation compares well: similar or better hourly rates, no vehicle costs, no mileage wearing down your car, no weather, and a fixed schedule in place of surge chasing. The phone-only tier at $10 to $15 an hour compares badly once you count unpaid upload time, and it wins only on flexibility and on not needing a car.
Against data annotation and labelling work, this is the same industry one layer down. Annotation pays less per hour at entry level and is easier to automate, since a model can increasingly pre-label and leave a human to confirm. Physical data collection resists that, because nobody can synthesise the original recording of a hand doing a thing. If you are choosing between them for durability, collection is the safer side of the same bet.
Against skilled remote freelancing, this loses on ceiling and wins on entry. You need no portfolio, no niche, no proposal and no client relationship to start, and that is rare. But no version of this work lets your rate climb with reputation the way a specialist's does, so it serves you best as a bridge to somewhere else.
Against a conventional part-time job, the differences that matter are classification and stability. You carry your own tax, you get no sick pay, and the contract ends when the dataset fills. In return you get schedule control at the remote tier and, at the on-site tier, a rate above most hourly work you can get without credentials.
The clearest way to hold it: this is well-paid unskilled work in a field that currently needs more hands than it has, and that will not last forever. Use it for what temporary conditions are good for, cash now and a foot in a growing industry, and keep your plans for 2030 elsewhere.
Rookie Mistakes
Treating the posted rate as your rate. The posted rate applies to accepted work. Rejections, setup time between takes and unpaid reading of the spec all sit between the two numbers. Track your real hours against your real earnings for the first two weeks and you will know which tier you are actually in.
Batching before validating. Recording forty clips against a spec you have misread turns a day into nothing. One clip, submitted, confirmed, then batch.
Performing for the camera. New collectors perform the task, making it smooth, readable and unnaturally tidy. Some specs want exactly that, and some explicitly want natural movement including fumbles and corrections, because a model that has only seen flawless demonstrations cannot recover from error. Do what the spec says, even when it looks worse on camera.
Ignoring the physical demands of the on-site tier. An eight-hour shift standing and moving through a facility, hitting productivity targets, four possible shift slots including a midnight to 6 AM crew, is industrial work. The listing says so plainly. People apply picturing a desk and a joystick.
Letting a project end without a next one. The Argentina listing runs roughly four weeks. That is normal: datasets fill and the work stops. Line up your next platform while the current contract runs, and keep two or three relationships alive at once.
Skipping the classification question. These are contractor roles. Nobody is withholding tax for you, and depending on where you live you may owe quarterly payments and self-employment contributions on this income. Budget for it in month one so year end holds no surprise.
Gotchas Worth Knowing Before You Apply
This is a job, with no asset at the end. You trade hours for money and the dataset belongs to the buyer. Nothing compounds, you hold no equity in what you built, and no residual arrives when the model trained on your data ships. That is a fair reason to do it, and a poor reason to expect it to grow into something else.
Geography gates the good tier. The $30 to $55 range is attached to on-site work in a US lab. If you do not live within commuting distance of one, that tier is closed to you whatever your skill, and the remote tiers pay a third as much. This is the single biggest factor in what this work can be worth to you, and your postcode decides it.
The night shift is real. A 24/7 facility staffs an 11:45 PM to 6:00 AM crew. Those hours are often easier to get, and they carry a real health cost that a slightly higher rate does not make up for.
Your home becomes the set. Phone-based capture means recording inside your kitchen and living space, over and over, to spec. The people you live with wander into frame and have to be kept out of it. Some people find this fine, and some find it quietly unbearable after two weeks.
Nondisclosure is standard and enforced. You will likely sign an agreement covering the client, the hardware, the specs and the tasks. It limits what you can say publicly, which makes the field hard to research, which is part of why so much of what you can find about rates is secondhand.
The spec can change mid-project. A buyer refining what their model needs will revise requirements, sometimes undoing a capture approach you had polished. Ask how revisions are handled and whether work already accepted stays accepted.
How would the people you live with feel about a camera running in the kitchen three evenings a week? Ask them before you apply, since their patience is part of your setup.
What You Are Signing Away
An offer in hand changes what matters. Before you accept one, look at what leaves your hands along with the recording.
The paperwork on this work deserves more of your attention than a $12 an hour job would normally get, because what you hand over is unusual.
You are recording the inside of your home, your hands, your movement patterns, and often your voice. Those recordings go into a training set, and training sets are copied, kept, resold between companies and used to build models that outlive your contract by years. You have no practical way to get your footage back later. Assume anything you record is permanent.
Four questions are worth asking before you sign, and a legitimate buyer will answer all four.
What is the scope of the licence you are granting? Most agreements take a broad, perpetual, worldwide licence to use the captured data for model training. That is normal in this field. What varies is whether it extends to your likeness for other purposes, and whether they can sublicense to third parties. Read those two clauses twice.
Who else appears in your footage, and did they agree? Housemates, family, children, and anything visible through a window. Most specs ban other people in frame, partly for data quality and partly because the buyer wants no consent problems. That ban protects you too, and breaking it can void a batch after you have already been paid for it.
What is visible that you did not intend? Post, screens, prescription labels, keys, documents, a laptop showing an email. A first-person camera at counter height sees a lot. Clear your set the way a photographer would, every session.
How is payment tied to acceptance? Being paid for hours worked and being paid for accepted captures are very different deals. The OpenTrain teleoperation listing states compensation is "hourly and paid ... for the work you perform during scheduled shifts", which is the structure that favours you. Piece-rate acceptance models push the cost of an unclear spec onto you. Ask which one applies and get it in writing.
Would you be comfortable if a clip of your kitchen and your hands were still in a training set ten years from now? If the honest answer is no, stop at this section, and that is a perfectly sensible place to stop.
Turning It Into Something You Own
The honest limit of this work is that it is hours for money. There are three ways people turn it into something with more leverage, from easiest to hardest.
Move to quality control. Reviewing and checking other people's captures pays more than producing them, needs the spec fluency you build in your first months, and grows without your hands being in frame. This is the most reliable step up, and the one buyers most often need filled, because their bottleneck is reviewers. Collectors are easier to find.
Become a collection coordinator. Buyers who need coverage across many environments have a logistics problem: recruiting, briefing, spec compliance and quality across a scattered group. If you have done the work, can read a spec and can manage people, you are a natural fit, and this is where the work starts to look like a business with margin.
Run a collection site. The capital-heavy version: a space, several rigs, a small trained crew, and a volume contract with a buyer. This is a real business with real risk, it depends on a handful of customers, and those customers may bring the work in house at any time. Understand it as the ceiling of this path, and leave it out of your first-year plans.
What all three share is that your transferable asset is your fluency with specifications and your reliability. Your dexterity matters far less. Both of those are visible to a buyer only if you have a record with them. That is the strongest argument for taking even the $10 an hour tier seriously while you are in it: the work is worth little, and the record is worth something.
Moving into quality control changes who you are in this trade. You become the reviewer whose sign off a buyer waits for, the person whose name they recognise from months of clean captures. That record travels with you to the next buyer, and it is the first piece of a business you can call your own.
Which of those three steps fits the person you already are: careful checker, organiser, or owner? Choose one early, and steer your first months of capture work toward it.
Behind the Scenes: What a Shift Actually Involves
The on-site version is much closer to precision manufacturing than to gaming.
You arrive for a shift that overlaps the outgoing crew, because the facility never stops. You are assigned a cell: a robot arm or a humanoid, a set of objects, a task list. The task is narrow and repetitive on purpose. Pick up the mug, place it on the rack, in this orientation, thirty times, with the starting position varied, because a model trained on one starting position learns only one starting position.
You drive the robot through it, often through a control interface with imperfect force feedback, which is why the work tires you in a particular way: your hands know what the task should feel like and the robot does not pass that feeling back. Depth perception through a camera is worse than you expect. Your first day will involve knocking things over.
Between takes there is reset work. Objects returned to the start, the cell tidied, now and then a snag passed to a technician. Reset time is a real share of your shift, and how briskly you reset is much of what a productivity target measures.
Every so often a batch gets reviewed and some of it comes back to you. A gripper hid the object at the moment of contact. The demonstration drifted outside the workspace. The lighting changed when someone opened a door.
The remote version has the same shape at a smaller scale. Rig the phone, mark the floor, run the task, check the clip, reset the kitchen, run it again. The chore is the easy part. The hard part is doing it identically twenty times while a phone watches, and staying accurate on the twentieth.
Could you pick up the same mug thirty times in a row and still care about the thirtieth? That patience, more than any skill, is what this job pays you for.
Waiting costs you the next project wave and the place in line that comes with an early record. The bar to start is low and the paperwork is the hard part. Tonight, pick two postings, read both data clauses, and send one application you would be comfortable signing.
Where This Goes Next
Some reasoning about where this is heading. Treat it as reasoning you can check, with no promise attached.
The entry tier gets squeezed first. Phone-only egocentric capture is the least distinctive thing a person can supply, and it is the tier most exposed to synthetic data and simulation getting better. Expect $10 an hour to stay flat on paper and lose value after inflation, while the wearable and on-site tiers hold up better because the hardware in the loop is harder to fake.
Specification skill becomes what sets you apart. The collectors who last will be the ones who can read a demanding spec, hit it consistently, and flag unclear points before wasting a batch. That is a quality-control skill, and quality-control roles pay more than collection roles. Your realistic path out of this work runs through reviewing other people's captures.
Geography spreads, then concentrates. Remote tiers will keep opening to more countries, because buyers want varied environments and cheaper hours. On-site labs will cluster near the robot developers. The gap between the two tiers is more likely to widen than to close.
Contracts stay short. Dataset-shaped demand produces project-shaped work. Plan for a portfolio of platforms over a single job, and treat every contract as ending on schedule even when a manager hints otherwise.
The information problem stays. Nondisclosure agreements plus fast-moving rates mean the public record on what this work pays will stay thin and unreliable. That is exactly why the habit of reading the posting and ignoring the rumour will keep serving you better than any rate guide, this section included.
Projects come in waves, and each wave goes first to operators with a track record. That record starts the day you complete your first approved capture. Wait for the field to feel settled and you join it after the people who learned the rigs in the uncertain months.