The data your model
hasn't already seen.
Public data is exhausted and contaminated. fiund sources non-public audio, video and UGC directly from the people who own it — with signed AI-training rights and consent that travels with every file.
100%
Every asset carries a signed licence granting AI-training rights explicitly.
Zero
Nothing pulled from the crawl, re-licensed off a public set, or of unknown origin.
Per clip
Machine-readable manifest with rights reference and consent status on every asset.
48 hrs
Send requirements and get scoping, availability, and an honest read within two days.
By modality
Real-world data, licensed at the source.
Sourced from creators, studios, and archives — not scraped, not re-licensed. Every modality below links to specs, availability, and how it's cleared.
Conversational speech
Natural, unscripted, multi-speaker dialogue — interviews, calls, and conversations recorded in real conditions.
Read speech
Prompted or scripted speech recorded in controlled conditions with matching transcripts.
Sung audio
Sung vocal recordings with performer consent, for music and voice modelling.
Film & professional video
Professionally shot film, broadcast, and catalog footage available for AI licensing.
Creator & UGC video
Creator and user-generated video sourced directly from the people who filmed it, with consent.
Egocentric video
First-person video from head-mounted and wearable cameras, capturing real tasks from the doer’s viewpoint.
Motion capture
Marker and markerless motion capture from consented performers, delivered rig-ready.
Sensor & IMU
Inertial and device-sensor data streams paired with activity labels.
Agentic trajectories
Step-by-step records of humans completing digital or physical tasks, for training agents.
How teams work with us
Brief to delivery, rights first.
Filter the catalog and order self-serve, or send a brief for sourced-to-spec — either way the paperwork is done before anything moves.
Full schema on every dataset page.
Rights & provenance
Diligence-ready. Every single asset.
Training-data deals fall apart over provenance, not quality. Every row we deliver is backed by a signed agreement you can pull in diligence — here's how a delivery reads:
Self-serve or sourced to spec
Tell us what you need. We clear the rest.
Filter the catalog and submit a purchase order, or send a brief — modality, languages, conditions, volume — and get scoping within 48 hours.
Filter the catalog →Send a brief