Consent Before Camera for Data You Can Defend
TLDR. If your model learns from real people, you need to show who agreed, to what, and when, without passing personal data around. We collect only after signed consent, we keep identity in a private registry, and you receive participant codes only.
Signed consent before any capture
No camera records and no microphone opens before consent is signed. That part is boring on purpose. It gives you a clean chain you can show later.
Each participant agrees in writing to the purpose, the capture types, and handling terms. Your specification shapes those materials. If you restrict age range, attire, location, or use, we write it into the protocol.
On site, the team checks consent status before each session. Missing or incomplete consent means no collection. You never receive material that falls outside what was agreed.
What consent covers in practice
- purpose and allowed use for your task
- capture types in scope, from video and images to speech and sensor
- handling terms, retention, and delivery form
- limits you set in your specification
What you receive and what we retain
You do not need names, phone numbers, or ID copies to train. We keep those separate. You get codes that link files to consent status and to attributes you asked for.
| You receive | We retain separately |
|---|---|
| Participant codes, not identities | Signed consent forms and identity records |
| Files tied to codes | Access-controlled registry linking codes to consent |
| Requested attributes like speaker group or demographic band | Source personal details, never in delivery |
| Manifest with checksums for every file | Withdrawal log and export record |
This split lets you audit coverage and trace a file back to an active consent without touching personal information.
How withdrawal works before export
Participants can withdraw before export. We remove the affected files from the delivery set and update the manifest. Simple.
You receive only active consents at export time. The manifest lists exactly what shipped, with checksums per file, so your records match ours during review. After export, that manifest is your stable reference.
Kinda like plumbing. The pipes carry water, not the reservoir. You get what you need to build, we hold what must stay private.
Content rules enforced on set
You set boundaries. We apply them during collection.
Teams are briefed on clothing, backgrounds, visible documents or screens, language, and behavior for your task. Checks happen at capture, where a retake is cheap, not after delivery when fixes are costly.
Coverage follows your spec across modalities. Video from HD to 4K and all image categories are checked against your rules in indoor and outdoor setups. Speech covers English, Hindi, Kannada, Telugu, Tamil, Malayalam, Marathi, and Sanskrit, plus more on request, with prompts screened and off-script material flagged. Depth, LiDAR, sensor, radar, and multimodal captures are reviewed for incidental personal or sensitive detail.
Answers legal teams ask for
Do you collect before consent? No. Signed consent comes first, verified on site before each session.
Do we receive personal data? No. You receive codes, files, requested attributes, and a manifest. Identities stay in a private registry with controlled access.
What happens if someone withdraws? We pull their material before export and ship a manifest that reflects only active consents.
Can we set our own content and use limits? Yes. Your specification defines them and we write them into consent materials, briefings, and checks.
Start with a pilot
Send a short spec and we will return a pilot plan with consent scope, capture setup, and manifest format. Both sides evaluate fit before scaling.
