Per-use data licenses for faces, voices, and performances — each with a signed consent receipt, a provenance trail, and an indemnification reference. Scoped to model training, time-boxed, and revocable. When someone pulls out, they drop from the next package.
Every face, voice, and performance is licensed from a verified person on scoped, time-boxed, renewable terms — with a receipt your counsel and your model card can stand behind. People are people here, never assets: they choose the use, they're paid, and they can pull out.
Every license issues a signed receipt — member, scope, term, provenance, and an indemnification reference. The clean-data story your training pipeline has been missing.
A first-class model-training scope, not a marketing clause stretched to fit. Faces, voices, and performances, licensed for exactly the training use — and nothing past it.
When a person pulls out, the license is invalidated and they drop from the next dataset package. Revocation is real, and it's the contributor's to use.
Clean provenance isn't a label you add later — it's how the data is licensed in the first place. Here's what runs at each stage, and the consent landscape it's built for. Watch the ledger run underneath.
Before any contribution is licensed, we prove they're a real, single person — who chose the training use.
A per-use data license, scoped to the training task — with a signed receipt that travels with the data.
Licensed contributions package with a manifest. A withdrawal excludes a person from everything that comes next.
Designed to these regimes. Not a certification.
We'll walk one consented contribution from a verified person to a dataset manifest your counsel can read.