hi! — T5: AI Video, Licensed First

Your call — two decisions

Raw AI output carries no US copyright, so every client deliverable needs our own authorship layer; the trial is one month of bake-off seats, cancelled after. Both are argued below.

Contextthe legal fact, the license map, where your taste gate sits, and what the trial buys

The one legal fact that shapes everything

Raw AI output has no US copyright. The Copyright Office's 2025 report: no human authorship, no protection — prompts alone don't count, and registration requires disclosing the AI layer. So hi! can never convey exclusive ownership of an unedited generation. Every client deliverable needs our real authorship layer — edit, arrangement, direction — and the engagement language says which layer is which. That isn't a burden; it's "sell judgment at AI speed, never hide the machinery," written into IP law. Provenance watermarks (Google's SynthID, OpenAI's C2PA) point the same direction: disclosure is our brand, not our problem.

The license map

Client-eligibleConditionalFlag

Tool	Verdict	Why, in one line
Adobe Firefly (own models)	Client-eligible	Licensed training data, enterprise indemnity, already ours — the baseline. Caveat: partner models inside the Firefly app aren't under Adobe's standard.
Google Veo 3.1 (Vertex/API, GA)	Client-eligible	GA models sit under Google's generated-output indemnity; SynthID in every frame.
Runway (paid)	Client-eligible	Cleanest ownership stance in the field: paid outputs are yours, commercial use explicit from $15/mo.
OpusClip-class editing	Client-eligible	Different risk class: it edits footage we already own.
OpenAI Sora (paid)	Conditional	Ownership assigned and commercial on paid tiers — but C2PA reliability disputes and shifting business-tier terms; re-verify at trial.
Higgsfield (paid)	Conditional	Commercial on paid plans, but the terms push all rights risk onto us — an aggregator wraps other models' risk and hands it over.
Kling (direct)	Flag	Grants a usage right, not ownership — internal and mood work only until read line-by-line for a specific campaign.
Free tiers, all vendors	Flag	No commercial rights. Nothing client-facing, ever.

Where your taste gate sits

Brief → asset direction (prompts, reference frames, negative space, brand hex) — Cairo
Generation batch (license-eligible tools only for client work)
Cairo first-pass cull — thirteen principles + the AI-tell audit; most generations die here
Elizabeth's taste gate — the culled shortlist on approval strips, notes on, small sets; client-usable vs internal-only assigned per asset
Human authorship layer — edit, grade, composite, type: what makes it ours and ownable
License log entry (tool, plan, date, indemnity) per claims-ledger discipline → delivery

The gate reviews shortlists, not firehoses, and sits before the edit investment — a generation costs credits, an edit costs hours. Which is also the verdict on "300 videos in 5 minutes": if the claim is real, it's a variants tool for internal ops, and precisely not taste-gated client creative.

The trial — what your approval buys

Seat	Cost	Why
Firefly video (own models)	$0 — ours	Sample 0, the license-clean baseline.
Runway Standard	$15/mo	The ownership benchmark.
OpusClip Starter	$15/mo	Repurposing, tested on our own footage.
Veo 3.1 via API	≤$20 cap	The indemnified GA benchmark.
Higgsfield Starter (optional)	$19/mo	Aggregator convenience vs its risk posture, tested once.

One month, one bake-off, one real brief, everything through your gate: $50–69. Losers get cancelled.

Phase 3 — the craft reference or dedicated skill — gets written after the bake-off, from our own verdicts on our own outputs. A reference built from vendor showreels would be the exact fluff this curriculum filters.