ditto.Join the waitlist
01 Robotics data collection

Teaching robots
the human hand.

Ditto is building motion-capture gloves at scale and the software that turns their raw signal into structured, verified training data. We are building the largest corpus of hand motion-capture data that has ever existed.

Dataset 1.0
Largest motion-capture corpus
Built for
Humanoid robotics
02 The data pyramid

Not all robot data is equal.

Robotics training data forms a pyramid. Each layer up is harder to gather but more useful as a control signal. Ditto operates at the top - the layer humanoid hands actually need.

0504030201
TIER 05
Motion-capture + pressure & heat
Availability
Hardest to capture
Value for robots
Highest
Full hand pose alongside contact, pressure, and thermal cues - the signal humanoids need to match the dexterity of a human hand.
WHERE DITTO OPERATES
03 The glove

Built to mass-produce.
Built for the field.

A motion-capture glove engineered for real-world collection - not a lab rig. Pull it on, switch it on, and go. Recording and parsing start automatically; there is nothing to configure.

01

Waterproof

Capture in any environment - rain, wash-down, or wet work. The signal keeps flowing.

02

All-day battery

Long sessions with no babysitting. Collect a full shift on a single charge.

03

Durable by design

Engineered for repeatable, real-world use across thousands of tasks.

04

Turn on & go

Recording and parsing begin the moment it powers on. Zero setup, zero configuration.

05

Every hand size

A full XS–L range so the fit - and the data - is right for every collector.

06

Pressure & heat sensing

Contact, force, and thermal cues at the fingertips - the detail robots need to truly mimic a hand.

Sized for every hand.

Available in a full size range, so the glove fits as well as the data it captures.

Extra smallXS
SmallS
MediumM
LargeL
04 4D point-cloud parsing

From raw stream to verified tasks.

Our task-segmentation algorithms parse the 4D point clouds coming off the glove - isolating meaningful actions and surfacing the insight buried in raw sensor input. No manual labeling, no annotation team.

01 · CAPTURE

Wear & record

Sensors stream hand pose, contact, pressure, and heat as the collector works.

02 · SEGMENT

Auto-parse

Task segmentation splits the continuous 4D stream into discrete, labeled actions.

03 · VERIFY

Confirm & ship

Parsed tasks are verified for completeness, then added to the training dataset.

05 The network

The world's largest motion-capture dataset for hands.

Thousands of collectors across companies and regions contribute every day, building the largest motion-capture corpus in humanoid robotics. Workers are paid by their companies for each verified task - once parsing confirms it is complete.

01

Collect with gloves

Workers perform everyday tasks while the gloves capture rich hand data.

02

Auto-parse & verify

Ditto segments and verifies each task automatically - no manual review.

03

Get paid per task

Companies pay their workers for every verified task once parsing is complete.

06 Get started

Bring Ditto to your world.

For companies

Turn everyday work into training data.

Equip your workforce with Ditto gloves and convert the tasks they already do into the most valuable data in robotics.

Contact us
For labs

Work with the Ditto dataset.

Building humanoid policies? Join the waitlist for early access to the world's largest motion-capture dataset for hands.