Five roles,one contract.
PRE-GEN works because each participant does one thing well. The protocol defines the rules. A registry operator runs the infrastructure. Copyright owners set the terms. AI providers ask before they generate. Creative studios get content with its rights attached. Here is exactly what each role does — and what it gets back.
The roles.
PRE-GEN
Role: the rulebook. Defines PG identifiers, the five signed objects (license, decision, receipt, evidence bundle, assertion), and byte-level test vectors every implementation must pass.
What it gives everyone: one contract. Any registry, any provider, any studio — same rules, same verification, no bespoke integrations. Open specification under Apache-2.0, owned by no one.
PRAMPTA
Role: runs the infrastructure. Verifies rights-holder identity, assigns PG codes, signs every allow/deny decision, keeps the hash-chained audit log, resolves disputes, issues provider credentials.
What it gives: you never run a server to use the protocol — rights holders register, providers get keys, studios get evidence. And the spec stays operator-agnostic: PRAMPTA is the first, not the only.
Actors, musicians, brands, estates
Role: the source of every permission. Register a subject once, set the terms — tiers, scopes, prices — or file a free unconditional opt-out.
What they get: every compliant AI asks before it generates. Authorized uses pay on their terms; unauthorized uses are provable. No monitoring burden — the ask happens automatically, at generation time.
Model & platform companies
Role: the ones who must ask. Image, video, and voice generators integrate the three-call flow — detect subjects in prompts, verify, keep the signed decision.
What they get: a legal shield three calls deep. Per-generation proof of authorization, structured refusals instead of silent risk, and an audit trail built for the regulator era.
Film, TV, advertising, games
Role: the ones who ship AI content into the world. Their pipelines call PRE-GEN before generating, collect the signed objects per scene, and export evidence bundles into production paperwork.
What they get: chain of title on autopilot. E&O insurers and distributors receive documentation, not promises — and content stays shippable as disclosure rules tighten.
Verifiers
Role: check the work. Platforms, insurers, courts, journalists — anyone can resolve a PG code at /lookup, verify signatures against public keys, and walk the audit chain offline.
What they get: trust that doesn’t depend on anyone’s word. The math and the audit chain are the witness.
How it works together.
One scene in a film, generated with a licensed actor’s likeness. Six moves, five signed objects, zero ambiguity afterwards.
Why PRE-GEN exists.
Generation got cheap. Permission didn’t. An AI-generated face, voice, or brand in a film is a legal event: someone owns that likeness, and “the model made it” has never been a defense.
Until now there was no standard way to ask, grant, and prove permission for one specific generation. Contracts live in inboxes. Consent disputes live in court. The burden of proof lands on whoever generated — months later, with nothing but a vendor’s word.
PRE-GEN makes permission a protocol: asked by machines, granted on the owner’s terms, verifiable by anyone — before the pixel exists, and for decades after.
Why the split between spec and operator matters.
PRE-GEN is an open specification; PRAMPTA is its first commercial registry. That separation is deliberate. If a studio’s legal department distrusts one operator, the protocol outlives that distrust — anyone can run a competing registry, and the signed objects verify the same way. Trust attaches to signatures and audit chains, not to a company’s word.
Take a seat.
Rights holders register free on PRAMPTA. Providers integrate via SDK. Studios start collecting chain-of-title evidence on the next production.