Real factory environments
Captured on live production lines, not in synthetic or staged setups.
Third Origin · Human workflow data
Train models on shopfloor reality.
Access the most exhaustive source of ready-labelled, end-to-end multimodal workflow data from diverse real-world manufacturing environments.
- Direct access to 200k+ manufacturers across industries
- Extensive ready‑labelled ego‑exo workflow data
- Custom data collection & tailored labelling based on your needs
Thirty task specimens · ego + exo · RGB–D · UMI bimanual · instrumented gloves · annotation demo · library map
01 · Platform
Real-world workflow data for physical AI.
Third Origin provides structured, training-ready workflow data from real manufacturing environments. Models trained on it see long-horizon tasks the way workers actually do them, decision points included.
End-to-end workflow coverage
Multi-step tasks with their transitions and decision points, not isolated pick-and-place.
Training-ready structure
Labeled, formatted, ready to load. Raw video alone does not train anything.
Model impact
Policies train on the mess they will actually meet, which shrinks the sim-to-real gap and holds up over long horizons.
02 · Task specimens
Task specimens.
Thirty tasks from working factories across a dozen sectors. Each note says what makes the task hard for a robot to learn.
01 glass works
Wrapping glassware for the box
Corrugated sleeves over fragile stock; grip force gets no second try.
02 appliance line
Fitting parts on an appliance line
Small device, moving belt, both hands busy at once.
03 assembly bench
Marking rings at the bench
Identical parts, tiny registration marks, zero tolerance for drift.
04 industrial qc
Counting impellers against the sheet
Count, check, log: perception and bookkeeping in one loop.
05 garment line
Trimming threads on knitwear
Snips against soft knit; one wrong cut shows forever.
06 knitting line
Feeding the knitting machine
Yarn tension by feel, machine timing by ear.
07 packaging line
Folding cartons for pouches
Flat blank to box in seconds, creases from muscle memory.
08 ceramics works
Finishing ceramic ware
Unfired clay marks at a touch; pressure is the whole skill.
09 electronics bench
Calibrating a camera module
Checkerboard, module, micro-adjust: sub-millimetre alignment, on camera.
10 pleating bench
Pleating fabric with a blade
A straightedge, a knife, and folds that must repeat exactly.
11 moulding bench
Filling a moulded tray
Each cavity takes one part, seated flush or not at all.
12 assembly line
Seating strip into extrusion
Long flexible part, narrow channel, continuous press-fit.
13 lampshade bench
Folding a pleated shade
Accordion folds that collapse if the order is wrong.
14 wax works
Banding wax blocks
Rubber bands around soft blocks: deformation on every grip.
15 garment line
Folding menswear to spec
The same fold every time; the pile is the QC.
16 leather works
Skiving leather at the bench
Blade angle by feel; one pass decides the edge.
17 garment line
Threading cords through hoods
A drawstring through a channel, blind for most of the run.
18 electronics bench
Testing a handheld unit
Cluttered bench, live device, probe and screen in parallel.
19 footwear line
Pressing shoe insoles
Layered soft parts aligned by hand before the press.
20 stainless works
Deburring under the machine
Gloved work under a running head; force where the camera can't see.
21 upholstery bench
Fitting a leather panel
Pry, stretch, seat: a stiff panel into a tight housing.
22 lockworks
Sorting padlock parts
A bowl of near-identical brass parts, picked by touch.
23 appliance assembly
Wiring appliance lids
A harness routed through housings, connectors seated blind.
24 cleanroom lab
Pipetting in the cleanroom
Gloved pipette work into microtubes; contamination is failure.
25 metal shop
Forming sheet metal
Rivets and sheet stock: high force, exact placement.
26 glass works
Bubble-wrapping washed glass
Wet, clear, slippery, and invisible to depth sensors.
27 packing floor
Strapping a bundle
Tension a strap around a soft bundle without crushing it.
28 footwear line
Bonding sole to upper
Machine-paced stitching where two curved parts meet.
29 print shop
Collating and cutting pages
Ruler-true cuts on stacks that shift as they square.
30 garment line
Attaching buttons and zips
A press machine, a placement mark, and no undo.
03 · What ships
Workflow Intelligence Layer.
The data foundation for physical AI: real tasks and real decisions, captured at scale.
Real-world task diversity
5,000+ real manufacturing tasks, with the natural variation and edge cases that make generalization possible.
Hierarchical action understanding
Data includes atomic actions and high-level human commentary, so one dataset teaches low-level control and task-level reasoning.
Multi-modal interaction data
Synchronized ego + exo video, tactile, and contact signals for learning grounded physical interaction and fine-grained manipulation.
Training-ready format
Ships as structured datasets with temporal alignment and segmentation, ready to drop into a training pipeline.
04 · Reach
Built on real manufacturing environments.
200,000+Manufacturers accessible
310+Manufacturers accessed
100+Distinct workflows captured
100,000+Hours of ready-labelled data
We work with factories that let us put cameras directly on the line: garments, cosmetics, packaging, electronics, mining, industrial tooling. Each sector brings its own tools, its own materials, and its own ways of going wrong.
05 · Diversity
Every point leads back to a specimen.
Explore 900 captioned windows from the same 30 hand-selected tasks shown in the carousel and specimen catalog. Search for a manipulation, inspect the cluster, then open the published clip that produced it.
0 / 900 moments · 30 source specimens
t-SNE projection of four-second window captions. Axes carry no units; nearby points describe similar work.
06 · Annotation · 190 steps · 21-point hands
Annotation, running live.
A two-minute ironing line with every step labelled and per-frame hand keypoints. Step boundaries come from the hand kinematics, not the model; the curve on the timeline is the motion signal itself.
00:00.00
Annotation timeline
0s10s20s30s40s50s60s70s80s90s100s110s120s
07 · Why us
Why Third Origin?
One team from spec to delivery
Define the spec with us, see early captures, and adjust before the full collection runs.
Built with robotics labs
Robotics and world-model teams shape what we capture, so the data answers training questions, not filming ones.
Real-world workflow data
Captured directly from live manufacturing environments: real tools, materials, constraints, and edge cases.
Workflow-driven data design
Task structures, action hierarchies, and annotations defined with partners, aligned to model training objectives.
Validated before it ships
Every modality passes multi-stage QA before delivery. If a stream is unusable, you never see it.
Secure and compliant data handling
Every dataset is rights-cleared and auditable, and we adapt handling to your compliance requirements.
08 · Products
Train models that work in the real world.
Off-the-shelf labeled datasets
Large-scale ready-labelled manufacturing workflow data
- Ego + exo capture
- Structured and training-ready
- Non-exclusive access
- Suited for robotics labs and vision / world-model teams
Custom data collection
Bespoke workflow capture
- Customer-specific task selection
- Tailored labeling and formatting
- Exclusive curated datasets for advanced teams
Task specimens
Thirty tasks across glass, ceramics, garments, electronics, footwear, metal, and packaging work, each with a note on why it is hard for a robot.
Watch the clips →Five annotated workflows
Filter assembly, ironing, arm assembly, parcel labeling, and pin insertion: 789 labelled steps with per-frame hand keypoints, playing in sync.
Open the viewer →The library, mapped
900 captioned windows from the 30 curated task specimens, searchable by what the hands are doing and traceable to the published clips.
Search it above →Third Origin begins where digital records end.
Workflow data that teaches models whole jobs, not isolated actions.