Egocentric training data
Ethically-sourced egocentric data at scale.
First-person, head-mounted capture — fully consented, license-clean, and model-ready. The training data AI labs can actually use.
What we deliver
First-person data, built for training.
Not stock footage and not scraped clips — purpose-captured egocentric data, structured for how models actually learn.
First-person video
Head-mounted, point-of-view capture of real tasks, hands, and environments — the perspective embodied models need.
Multimodal signals
Synchronised audio, IMU motion, depth, and gaze captured alongside every frame of video.
Frame-level annotations
Actions, objects, and interactions labelled and quality-checked, delivered in the schema you train on.
How it works
Collect. Consent. Annotate. Deliver.
A pipeline designed so the data arrives clean, documented, and ready to train on.
01
Collect
Contributors capture first-person footage of defined tasks and scenes.
02
Consent & QA
Every contributor is consented; footage is screened for quality and PII.
03
Annotate
Frames are labelled for actions, objects, and interactions, then reviewed.
04
Deliver
Clean, documented, model-ready datasets in your preferred format.
Provenance & ethics
Data you can actually license.
In a market full of legal landmines, clean provenance is the product. Every dataset is consented, documented, and defensible.
Consent on every contributor
Each person who captures data signs a clear, documented consent and licensing agreement.
Clean, traceable licensing
Every dataset ships with provenance you can audit — no grey-area scraping.
PII & face handling
Faces and personal information are screened and handled to defined privacy standards.
IP indemnity
Licensing structured so you can train with confidence, not legal exposure.
Use cases
Where first-person data moves the needle.
Need a dataset that doesn't exist yet?
Tell us the task, the environment, and the volume — we'll commission the capture.
Samples
See the data before you commit.
Sample audio datasets
A representative slice of our audio data, available under request access.
Request accessEgocentric datasets at scale
First-person, head-mounted datasets are in active collection. Talk to us about early access and bespoke commissions.
Get on the listSecurity & compliance
Handled like the asset it is.
Let's get your model the data it needs.
A 20-minute call to scope the data, the consent, and the timeline.