One row, four named services
The catalog row names lash extension, tinting, shaping and brow styling and nothing else. A spec says which of the four your policy needs and where each one starts and ends.
Now recruiting capture teams worldwide. Get paid to record first-person video of real work.
Apply to captureHard task
Requirements
The catalog row names lash extension, tinting, shaping and brow styling and nothing else. A spec says which of the four your policy needs and where each one starts and ends.
The catalog lists a head-mounted phone or a multiview rig. A head-mounted view keeps the practitioner’s hands and the tool in frame, and a multiview rig adds synchronized angles when the spec asks for them.
3D hand pose carries a per-joint visibility flag, so a finger folded behind a small tool is marked as hidden rather than silently filled in.
The site states that faces are irreversibly blurred at ingest and that consented subjects are exempt. Whether the person receiving a service is in frame is set in the skill spec.
Every episode links to a signed contributor release, and the license is perpetual and buyer-owned.
Definition
It is first-person video of the work named in one catalog row: lash extension, tinting, shaping and brow styling services. A practitioner does the work in a salon or studio while a head-mounted phone or a multiview rig records it, with hand pose and sensor streams alongside. People reach the same request through several phrasings: egocentric eyelash extension data, first-person brow shaping and tinting video, a beauty service robot dataset or custom salon data collection at scale. They describe one request from different angles.
The data is human capture by paid, consented contributors. It is not robot or teleoperation data, and this page makes no claim that any robot has been trained on it or that any robot performs these services. It describes task video only. The site makes no health, allergy, cosmetic-safety, licensing or regulatory claim for this row, and nothing here is guidance on how to do any of the four services.
Catalog row
One row. "Eyelash & Eyebrow Services" sits in the Hair Salon, Barbershop & Beauty Studio environment, in the Beauty & Personal Care sector. It reads: "Lash extension, tinting, shaping and brow styling services." It is rated Hard, lists "Phone (head-mounted) / multiview rig" as the capture setup, and is Available on request. The first table below reproduces it as the catalog states it. The environment’s shared examples line reads "Hair salon, barbershop, nail studio, beauty spa -- cutting, coloring, styling, nail treatments, workspace sanitation", and it is one line for every row in that environment, so it is quoted here and not assigned to this row alone.
Available on request means the work is scoped and recruited to spec when a buyer asks. It is a statement about the catalog row, not a promise of a delivery date, and it does not mean a stock of lash or brow footage exists. This page claims no volume, hours or delivered footage for the row. The row names four services. It names no product, no tool, no technique and no outcome, so if your program needs any of them, they are written into the skill spec and not assumed.
Why it is hard
The site’s answer is the Skill column. The catalog rates this row Hard, and that is all "Hard" means here. This page imports no outside rubric and states no failure rate, tolerance or claim about how skilled any worker is. What the site does say sits in the row text and in its other beauty copy. The row bundles four different services, so one row covers four kinds of movement. The site’s stereo makeup page describes the row as the finest work in its pair of beauty rows, with small tools, small parts and a hand held close to the work. The site also states that the hands occlude the work and that fingers disappear behind objects.
The site publishes no step list, no duration and no reading for any of the four services. The failure-mode table below maps each difficulty to the stream or label the site already states, and says Not stated where nothing is. The neighbouring Makeup Application and Basic Skin & Beauty Treatments rows are rated Medium, so this is the one Hard row of the three on the egocentric makeup page.
Difference
The egocentric makeup and lash application page covers three Available on request rows from the first-person side: Makeup Application, Eyelash & Eyebrow Services and Basic Skin & Beauty Treatments. Makeup application leads that page. This page takes only the Hard row, so it is the page to start from if your program is about lash and brow work and not about makeup or skin treatments. It does not replace the makeup page, which stays the home for the other two rows.
The stereo makeup and lash application page covers two rows, Makeup Application and Eyelash & Eyebrow Services, with a calibrated stereo pair and MCAP delivery. This page is the first-person treatment of the single lash and brow row and does not describe stereo capture as running on it. The egocentric hair salon and beauty page lists all 13 rows in the environment, and the stereo hair salon and barbershop page covers hair, nails and sanitation rows and does not cover makeup, lash or brow work. The second table below sets these pages side by side.
Services
The row text names lash extension, tinting, shaping and brow styling. That is the whole of what the site states about them. Which products, tools, lash types or brow techniques are involved, how long a service takes and what a finished result looks like are Not stated. The site lists no outcome or quality label for the row, so what counts as a finished service is set in the skill spec.
A spec that says "do eyelash and eyebrow services" gives one long action segment. A spec that names each service, and the steps inside it, gives a sequence you can slice, because action segments carry start_ns, end_ns, verb and noun. Whether an episode covers one service or a run of several is a spec decision, since the site sets no episode length. The third table below sets each named service beside the streams and labels the site states.
Capture setup
The catalog lists Phone (head-mounted) / multiview rig for this row. The site describes two capture modes across its beauty and other commercial rows. In the first, a head-mounted phone records the first-person view of the task. In the second, a multiview rig adds synchronized third-person angles when the spec asks for them. The row lists both, and which one an order uses is settled in the skill spec. The site also states that Firsthand does not capture at a customer’s own facility, so the setting is agreed in the spec.
The site does not state a stereo pair for this row. Stereo applies only through the general stereo offer, agreed in the spec and confirmed on a Spec Pilot, and the stereo makeup and lash page is where that option is covered. The site also does not state a working distance or how close-range work is handled for this row, and it does not state lighting. Settle all three in the spec and on a Spec Pilot.
Streams
Firsthand custom collections record ten hardware-synchronized streams aligned under 2 ms. They include head and wrist RGB, depth, 3D hand pose for both hands with a per-joint visibility flag, contact and grasp events, and frame-accurate action segments. Hand pose is 21 joints per hand, 3D, in metres in the head-camera frame. The one language label the site states is a task instruction for each episode. Standard capture is cameras, depth and IMU. A separate glove tier records a tactile signal, and the site does not state whether it applies to this row.
Contact events mark when a hand meets a tool or a surface, and per-frame instance masks are stored at the depth resolution of 848×480. Which objects get masks and names, such as the tool or the work surface, is set in the spec. The site lists no label for the outcome of a service, such as the look of a brow or the state of a lash set, and none is claimed here. The streams are described for custom collections in general. Row-by-row stream coverage for each capture mode is confirmed in the spec.
Faces and consent
This row is performed on a client’s face, so the face rule matters. What the site states is general. Every episode links to a signed contributor release through a documented consent chain, and the consent record is part of the delivery. Faces, plates and screens are detected and irreversibly blurred at ingest, consented subjects are exempt from blurring, and a reviewer confirms the result before anything ships. Footage with bystanders who cannot be de-identified is discarded, and the anonymization record ships with the data.
What the site does not state is whether the person receiving a service is in frame for this row, whether they sign a release, or how their face is handled beyond that general rule. This page therefore does not say that identifiable client faces are, or are not, in the delivered video. Settle it in the skill spec before capture and confirm it on a Spec Pilot. Container labels and screens are not listed as blur categories beyond the screens the site names. This is not legal advice, and the executed agreement controls.
Visibility
The site states that the hands occlude the work and that fingers disappear behind objects. That is why every joint carries a visibility flag, why contact events mark when the hand meets a tool, and why instance masks are recorded per frame. It also states that reach and grasp frames are where blur and occlusion show up, and that depth degrades on wet and reflective surfaces, with a confidence map delivered so low-confidence regions can be masked.
The site states that metric depth on sub-centimetre parts is flagged rather than interpolated. Whether a single lash or a brow hair counts as such a part is Not stated, and so is a mask accuracy figure for them. A spec that needs a clear view of a step should name it, ask for wrist views or a multiview angle, and confirm it on a Spec Pilot.
Validation
Validation runs against signed accept criteria written into the spec before capture, and a reject log ships with each batch. Calibration is solved at the start and end of every session under a 0.28 px RMS reprojection gate. Episodes are machine-checked against stated reject criteria, such as a sync breach beyond the 2.0 ms ceiling for more than three consecutive frames, the manipulated object out of frame for more than 15 percent of the episode, or more than 35 percent of depth pixels in the manipulation region invalid, and a human reviewer then checks each one. Only footage that clears the criteria is billed as a validated hour.
For this row, the spec has to say which item counts as the manipulated object at each step and where the manipulation region is, because the tool, the work surface and the area being worked are different things at different moments. What counts as a finished service is also written down, since the site states no outcome definition. Rejected items are recaptured at the provider’s cost, and the site states no accuracy figure for step labels, so what counts as a pass for a step boundary is set in the accept criteria.
Delivery and license
Delivery is in the format your stack already reads: RLDS, LeRobot, WebDataset, HDF5, zarr or Rerun .rrd, or raw streams with metadata, with readable converter source alongside. Firsthand also offers MCAP delivery, an open source container for multimodal log data, and it can be the only delivery or sit beside one of the others. The site states no MCAP topic names, message schemas or rates for this row, so they are agreed in the spec. The fifth table below sets each format beside what its official documentation states.
The license is perpetual, irrevocable and worldwide, buyer-owned and non-exclusive by default, with exclusivity available as a separate term. The executed agreement controls. Because the data is human capture, using it for a specific robot still needs your own retargeting step. Official documentation for the formats is linked in the sources below.
Speed and scale
The October 2026 off-the-shelf catalog lists ~324,800 hours across 18 datasets. The site names no eyelash or eyebrow dataset among them, so this work is custom collection, not an off-the-shelf purchase. It is custom salon data collection at scale through a network of thousands of capture providers, for a consistent, recurring data flow to labs.
From signed spec to first validated batch, the published timeline is a median of 48 hours where coverage is deep and three to four weeks where it is thin and contributors need to be recruited first. The row is Available on request, so the range that applies is confirmed when the spec is quoted. After a first batch, on-demand top-ups run against the same accept criteria. Collections are billed per validated hour, and rejected episodes and recaptures are never billed.
Ordering
Describe which of the four services you need, where each step starts and ends, what you expect to see as the tool and the work surface, whether an episode covers one service or a run, what counts as a finished step, whether the person receiving the service is in frame and how that is handled, and the setting and lighting. Then name the capture mode, head-mounted phone, multiview rig or both, any wrist views, and the delivery format.
Most programs begin with a Spec Pilot, a small, fast first batch that proves the spec and the accept criteria on real footage before volume scales. The last table below shows what each spec element decides. Our skill spec guide walks through how to write one.
At a glance
Every collection runs through the same seven-stage end-to-end custom collection pipeline. End-to-end custom data collection is a managed service that takes an AI data need from problem to owned dataset in a single accountable pipeline.
Live catalog row
| Task | Description | Skill | Capture | Status |
|---|---|---|---|---|
| Eyelash & Eyebrow Services | Lash extension, tinting, shaping and brow styling services. | Hard | Phone (head-mounted) / multiview rig | Available on request |
| Makeup Application (neighbouring row) | Professional makeup application in a salon or studio | Medium | Phone (head-mounted) / multiview rig | Available on request |
| Basic Skin & Beauty Treatments (neighbouring row) | Facial cleansing, moisturizing, exfoliation, foot spa and waxing treatments | Medium | Phone (head-mounted) / multiview rig | Available on request |
Compared pages
| Page | What it covers | Where lash and brow work appears | Relation to this page |
|---|---|---|---|
| This page: egocentric eyelash and eyebrow services data | The single Eyelash & Eyebrow Services row, first person | The whole page | The dedicated page for the Hard row |
| Egocentric makeup and lash application data | Three rows: Makeup Application, Eyelash & Eyebrow Services, Basic Skin & Beauty Treatments | One of three rows; makeup leads the page | Broader first-person page for the neighbouring rows |
| Stereo makeup and lash application data | Two rows, Makeup Application and Eyelash & Eyebrow Services, with a calibrated stereo pair and MCAP | One of two rows | Stereo and MCAP treatment of the pair |
| Egocentric hair salon and beauty data | All 13 rows in the environment, first person | One of 13 rows | Environment-wide page |
| Stereo hair salon and barbershop data | Hair, nail and sanitation rows in stereo | Not covered | Neighbouring rows, not lash or brow work |
| Hard-task hub | Every catalog row rated Hard, in tables | One row in a table | The hub row links here |
Services and streams
| Service named in the row | What a skill spec can separate | Streams and labels the site states | What the site does not state |
|---|---|---|---|
| Lash extension | Its steps as their own action segments | Hand pose with visibility flag, contact events, action segments | Products, tools, lash types, technique, outcome |
| Tinting | Its steps as their own action segments | Hand pose with visibility flag, contact events, action segments | Products, technique, outcome |
| Shaping | Its steps as their own action segments | Hand pose with visibility flag, contact events, action segments | Tools, technique, outcome |
| Brow styling | Its steps as their own action segments | Hand pose with visibility flag, contact events, action segments | Tools, products, technique, outcome |
| Any of the four, as one episode | One service or a run of several | Task instruction, one per episode | Episode length, step lists, durations |
Stated vs not stated
| Item | What the site states | For an eyelash and eyebrow order |
|---|---|---|
| Catalog capture setup | Phone (head-mounted) / multiview rig | Which one an order uses is set in the skill spec |
| Stereo pair | Not stated for this row; available through the general stereo offer | A spec choice, confirmed on a Spec Pilot |
| Streams | Ten hardware-synchronized streams aligned under 2 ms; 21-joint hand pose with per-joint visibility flag; contact events; action segments | Stated for custom collections; row-by-row coverage is confirmed in the spec |
| Synchronization and calibration | Hardware-triggered onto one clock, 2.0 ms ceiling; calibration solved every session, 0.28 px RMS gate | Stated |
| Faces and consent | Signed release per episode; faces irreversibly blurred at ingest; consented subjects exempt; reviewer confirms | Stated as a general rule |
| Whether the person receiving the service is in frame | Not stated | Set in the spec before capture |
| Products, tool models, technique labels | Not stated | Not claimed; set in the spec |
| Outcome or quality labels for a service | Not stated; no such label is listed | Not claimed; set in the spec |
| Working distance, lighting, close-range depth behaviour | Not stated | Settled in the spec and on a Spec Pilot |
| Row status | Available on request | No volume, hours or delivered footage is claimed on this page |
| Health, allergy, cosmetic-safety, licensing or regulatory claims | Not stated | Not claimed on this page |
| Cycle time, throughput, accuracy, error rate | Not stated | Not claimed on this page |
| Force or tactile data | A glove tier records a tactile signal; whether it applies to this row is not stated | Set in the spec, if needed |
Failure modes and QA
| Failure mode | What the site states | What a buyer sees |
|---|---|---|
| Fingers hide behind a small tool | Per-joint visibility flag on hand pose; per-frame instance masks | A hidden joint flagged as hidden, not guessed |
| The work area leaves the frame | Episode rejected when the manipulated object is out of frame for more than 15 percent of it | Reject log per batch; what the object is gets set in the spec |
| Sub-centimetre parts | Metric depth on sub-centimetre parts is flagged rather than interpolated | Whether a lash or a brow hair counts: Not stated |
| Weak depth on reflective surfaces | Depth degrades on wet and reflective surfaces; confidence maps mask low-confidence regions | Behaviour on this row’s surfaces: Not stated |
| Invalid depth in the manipulation region | Episode rejected when more than 35 percent of depth pixels are invalid | Reject log per batch; confidence maps ship with depth |
| Judging a finished service | No outcome or quality label is listed | Not claimed; set in the spec |
| Bystander in frame | Faces and screens blurred and reviewer-confirmed; footage that cannot be de-identified is discarded | Reject log per batch; anonymization record ships with the data |
| Calibration drift | Solved at the start and end of every session; 0.28 px RMS gate | Calibration shipped with the episode |
| Sync breach | Rejected beyond the 2.0 ms ceiling for more than three consecutive frames | Reject log per batch |
| Spec mismatch | Rejected for a mismatch against the named skill spec | Reject log per batch |
| Grip and handling force | No force or tactile stream is stated for this row | Not claimed on this page |
Formats and licensing
| Format or term | What the official documentation or site states | Used for here |
|---|---|---|
| RLDS | A dataset is retrieved as episodes, and each episode holds a dataset of steps (RLDS repository) | Episode-based training loaders |
| LeRobot | LeRobotDataset v3.0 is a standardized format for robot learning data, including multi-camera video (LeRobot documentation) | LeRobot dataset format and tooling |
| WebDataset | Tar files in which files that belong together share a basename (WebDataset repository) | Sharded streaming for large training runs |
| MCAP | An open source container file format for multimodal log data, supporting multiple channels of timestamped pre-serialized data (MCAP site) | Optional container; topics and schemas agreed in the spec |
| HDF5, zarr, Rerun .rrd | Firsthand only | Array storage and synchronized playback for review |
| License | Perpetual, irrevocable, worldwide, buyer-owned, non-exclusive by default; exclusivity is a separate term | Executed agreement controls |
Skill spec
| Spec element | What it decides | Where it shows up |
|---|---|---|
| Services | Lash extension, tinting, shaping, brow styling, or a subset, worded as the catalog words them | Set per program; Available on request as listed |
| Steps and order | What one step is, and where each starts and ends | Episode boundaries and action segments; set per program |
| Episode scope | One service or a run of several | Set per program; the site does not state an episode length |
| Object definition | Whether the tool, the work surface or the area worked counts as the object at each step | Contact events and the out-of-frame reject rule; set per program |
| Finished-step definition | What counts as a finished service or step | Accept criteria; the site states no outcome definition |
| Capture mode and streams | Head-mounted phone, multiview rig or both, wrist views, hand pose, IMU | The channels you can read; confirm on the Spec Pilot |
| Person receiving the service | Whether they are in frame and how that is handled | Set per program; faces blurred at ingest, consented subjects exempt |
| Setting and lighting | Where capture runs and how the space is lit | Set per program; the site states no lighting setup for this row |
| Delivery format | How the data is packaged | RLDS, LeRobot, WebDataset, HDF5, zarr, Rerun .rrd or MCAP |
| Accept criteria | What passes QA, including what counts as a reject | Reject log per batch |
| License | Who owns the dataset | Perpetual, irrevocable, buyer-owned, non-exclusive by default |
Example footage
One real clip already published on this site, shown as a format example: ten seconds of first-person home cooking with faces blurred. It does not show eyelash, eyebrow or any beauty service.
No eyelash or eyebrow clip is published on the site, so this is labelled as a format example. Request a sample to see delivered files. See sample data outputs, request an egocentric data sample.
FAQ
Sources
Tell us which of the four services you need, where each step starts and ends, what counts as a finished step, whether the person receiving the service is in frame, the capture mode, the setting and lighting, and the delivery format. We will quote reach, timeline and price against a written spec.