How to Structure an AI Vertical Drama Commission for Legal Review

How to Structure an AI Vertical Drama Commission When Your Legal Team Has Never Seen One

The commercial conversation finished three weeks ago. Scope agreed, budget agreed, start date pencilled in. Then the draft went to legal, and it came back with forty two comments, most of them variations on the same question: what exactly is being delivered, and who owns it. Nobody on the legal team has papered a commission like this before. There is no precedent folder to pull from. So the review stalls, the start date slips, and the production partner is holding capacity for a series that may never be signed.

This is the most common reason a first vertical drama commission dies. Not price. Not creative disagreement. Contract structure that nobody in the building knows how to write.

The fix is not a clever clause. It is sequencing. Legal review moves quickly when the commercial terms have already resolved the questions legal is going to ask. Most first time commissions arrive at legal with those questions still open, and legal is asked to close them in a redline, which is the slowest possible place to close them. What follows is the order that works, and the specific decisions each stage needs to make before the document leaves the commercial team.

1. Start With the Deliverable Definition, Not the Technology

The single most useful page in the entire agreement is the deliverable schedule, and it should be written before anything else. Not a description of the series. A description of the files.

Episode count, runtime per episode, aspect ratio, resolution, frame rate, container format, audio configuration, subtitle format and language, thumbnail and key art requirements, metadata fields. Then the delivery mechanism and the acceptance window. A legal team can paper almost anything once it knows what a completed obligation looks like, and it can paper almost nothing when the obligation is described as a finished series.

The mistake here is treating technical delivery specifications as an operational detail to be settled later. They are the definition of performance. If the delivery schedule says seventy episodes at ninety seconds, vertical nine by sixteen, 1080 by 1920, twenty five frames per second, stereo audio at minus sixteen LUFS integrated, with burned in subtitles in the primary language and separate sidecar files in two secondary languages, then acceptance is testable. If it says a complete vertical drama series, acceptance is an opinion.

Write the schedule with the person who will actually ingest the files. Not the person who commissioned the series. Platform ingest teams have specifications already, usually documented, usually ignored during commissioning. Pull them forward.

2. Rights: Separate the Series From the Assets Underneath It

Vertical drama commissions carry two distinct rights questions and most first drafts collapse them into one, which is where the redline explodes.

The first question is ownership of the finished series. That is conventional. Work made on commission, assigned on payment, with the usual carve outs for the production partner's pre existing tools and general know how. Nothing about the production method changes this analysis, and legal teams handle it comfortably once it is framed as a standard commissioned work.

The second question is ownership of the production assets underneath the series. Character reference sets, style references, trained character models, prompt libraries, scene templates, continuity documentation, the generation configurations that produce a consistent look. These are not the series. They are the means of making more of it, and their commercial value shows up eighteen months later when a sequel is commissioned or a spin off is planned.

Decide explicitly. Three positions are defensible. Full assignment of assets alongside the series, which costs more and is worth it if a franchise is likely. Retention by the production partner with a perpetual licence back to the commissioning party for use on that title and its derivatives. Or joint custody with an escrow arrangement, where assets sit with a third party and release on defined triggers.

What is not defensible is silence. A contract that assigns the series and says nothing about the assets leaves the commissioning party with a completed show and no ability to continue it without returning to the same partner on whatever terms are available then. That is a negotiating position, not an accident, and it should be a decision rather than an oversight.

3. Talent, Likeness and Voice

Every commission involving synthetic performance needs a clean chain of consent, and this is the clause legal will spend the most time on, correctly.

Establish where every human element originated. Performances captured from real actors under release. Likenesses licensed from a talent library with documented consent terms. Voice performances recorded under release, or synthesised from a licensed voice model with its own consent documentation. Composite or fully synthetic figures not derived from any identifiable individual.

For each category, the agreement needs a representation from the production partner about provenance, a warranty that consents are in place and cover the intended distribution territories and terms, and an indemnity that survives delivery. Ask for the underlying documentation as a condition of final payment, not as a good faith promise. The paperwork exists or it does not.

Where union agreements apply, they apply. The SAG-AFTRA framework for synthetic performers and digital replicas sets out consent and compensation obligations that a commissioning party inherits by distributing the result, and the current terms are published at sagaftra.org. A production partner working outside the United States may still be delivering into a market where those obligations attach. Ask the question in commercial negotiation, not in redline.

4. Build the Approval Gates Into the Contract

The most valuable structural feature of a well written commission is a small number of hard approval gates, each with a defined artefact, a defined review window and a defined consequence for silence.

Four gates cover most series.

Concept and script gate. The artefact is the series bible and the full episode outline, plus scripts for the first block. Approval here locks narrative structure and the paywall or cliffhanger placement.

Look development gate. The artefact is the character reference set, the style guide and a small number of finished sample shots. Approval here locks visual identity. This is the gate that prevents the most expensive category of dispute, because it moves the aesthetic conversation to a point where changing direction costs a reference set rather than a series.

First block gate. The artefact is the first block of finished episodes, delivered to full specification. Approval here confirms that the pipeline produces the agreed standard at the agreed quality, and it is the natural place to sit the largest payment tranche.

Final delivery gate. The artefact is the complete package against the schedule from section one.

Each gate needs a review window in business days and a deemed approval provision. Without deemed approval, a slow internal review becomes a production stoppage that the production partner has no contractual protection against, and the partner prices that risk into the fee. With it, the commissioning party carries the consequence of its own review speed, which is where the consequence belongs.

5. Quality Standards That Survive a Dispute

Acceptance criteria written as adjectives cannot be enforced. Acceptance criteria written as tests can.

Split the standard into three layers. Technical conformance, which is the delivery schedule and is pass or fail. Continuity conformance, which is measurable against the character bible and the continuity documentation: the character wears the same costume within a scene, the location matches its established reference, the timeline holds. Craft conformance, which is the layer people mean when they say quality, and the only layer that requires judgement.

For the craft layer, define the standard by reference rather than adjective. Nominate an approved sample, usually the sample shots from the look development gate, and define the standard as consistency with that sample. This turns an unbounded argument into a bounded comparison. It also gives the production partner something to work against, which raises the delivered result.

Add a defect tolerance and a remediation mechanism. A percentage of shots eligible for retake without additional charge, a defined turnaround for retakes, and an escalation path when the parties disagree about whether a shot is defective. Most disputes never reach the escalation path. Their value is that they exist.

6. Payment Tranches Tied to Gates, Not to Dates

Tie every tranche to an approved gate artefact rather than to a calendar date. Date linked payments create an obligation to pay for work that may not have happened. Gate linked payments create an obligation to review, which is the behaviour the commissioning party actually wants from itself.

A workable shape places a mobilisation payment at signature, a payment at look development approval, the largest payment at first block approval, and the balance at final delivery and acceptance. Hold a retention against the acceptance window. Release it automatically when the window closes without a defect notice, so that the retention does not become a mechanism for indefinite delay.

7. Change Control and Model Substitution

This clause has no equivalent in a traditional production agreement and it is the one most often missing from a first draft.

Generation tooling changes during production. A model version updates, a provider changes its terms, a capability that the pipeline depends on is deprecated. The commissioning party does not need approval rights over the production partner's tool choices, and asking for them creates a review burden nobody wants. What the commissioning party needs is protection of the output standard.

Write it as an outcome obligation. The production partner may change tools, models or versions at its discretion, provided the delivered material remains consistent with the approved look development sample and the technical schedule. Where a change would produce a visible difference in the approved look, the partner notifies and the parties agree an approach before that change reaches delivered episodes.

Pair this with a provenance and rights warranty covering the tooling itself: that the partner holds the necessary commercial licences for the tools used, and that the output carries no restriction preventing the intended commercial exploitation. Terms differ meaningfully between providers, and a consumer tier licence attached to a commercially distributed series is a real and avoidable exposure.

8. Termination, Handover and What You Get If It Stops

Ask the uncomfortable question during negotiation, because the answer is cheap now and expensive later. If this agreement ends at the look development gate, what does the commissioning party hold, and can anyone else continue from it?

Define a handover package that scales with the stage reached. At minimum: approved scripts and the series bible, approved character and style references, continuity documentation, delivered episodes in source and mastered form, and the metadata that identifies which asset produced which shot. Whether trained models and prompt libraries are included depends on the rights position taken in section two, and that is precisely why section two comes first.

Set a handover window in days and make final payment conditional on it. A handover obligation with no deadline is not an obligation.

9. Where Legal Review Usually Stalls, and How to Unblock It

Three points account for most of the delay.

Legal cannot assess an indemnity when it does not know what is being indemnified. The provenance table from section three resolves this, and it should be an annexe rather than a paragraph.

Legal cannot advise on acceptance risk when acceptance is undefined. The three layer standard from section five resolves this.

Legal cannot size the rights position when the asset question is open. Section two resolves this, and it is a commercial decision, not a legal one. Do not send it to legal to decide.

Give the review team the deliverable schedule, the provenance annexe, the gate structure and the asset position as a covering note. Reviews that arrive with those four items attached tend to close in a single pass. Reviews that arrive without them tend to close in four.

Axis AI Studios Perspective

Axis AI Studios is an AI native vertical drama production studio based in the Netherlands. We produce commissioned series end to end, and we work with commissioning teams that are structuring this kind of agreement for the first time.

Our position is that a production partner should make the contract easier to write, not harder. That means arriving with a delivery specification rather than waiting to be given one. It means documenting provenance for every human element in the series as a matter of course, so that the warranty a client needs is backed by files rather than assurances. It means proposing the gate structure ourselves, because gates protect both sides, and a partner who resists defined acceptance criteria is telling you something.

It also means being direct about the asset question. Character references, style guides, continuity documentation and generation configurations have value beyond the first series, and clients should decide what happens to them with full visibility rather than discovering the position when they commission a sequel. We would rather negotiate that in week one than litigate it in month eighteen.

We work with platforms, brands, IP holders and media companies commissioning original vertical drama. If you are structuring a first commission and want a delivery specification, gate structure and provenance annexe you can hand to your legal team, write to business@axisaistudios.com.

FAQ

Do we need a separate agreement for the AI production element, or does one contract cover it?

One contract covers it. Vertical drama commissioning is a production services and delivery agreement, and the production method sits inside it as a set of warranties, standards and change control provisions rather than as a separate instrument. Splitting it across two documents creates gaps between them, usually around acceptance and rights in production assets. Keep it in one agreement with the technical material in annexes, which is also easier to update between series.

How do we handle the situation where our platform requires rights we cannot get from a talent library?

Establish the distribution requirement first, then work backwards to sourcing. If the platform needs perpetual worldwide rights across all media and the available likeness licence is territory limited or term limited, the production partner needs to know before look development begins, because the answer changes which sources are usable. This is a commissioning brief question rather than a contract question, and resolving it late means rebuilding character references that have already been approved.

What happens if the production partner uses a tool whose terms change mid production?

The change control provision should place that risk with the production partner as an outcome obligation. The partner is responsible for delivering material that meets the approved standard and carries no licensing restriction on commercial exploitation, whatever tooling changes occur along the way. The commissioning party should require notification where a change affects the approved look, and should hold a warranty on tooling rights that survives delivery, but it should not take on approval rights over the toolchain itself.

Further Reading

For the pre signature preparation that sits upstream of the agreement itself, the vertical drama commission checklist covers the commercial, creative and rights decisions a business needs to have made before a draft is worth circulating.

For the operational standards that turn an acceptance clause into something enforceable, what an AI vertical drama production SLA should actually contain sets out the specific provisions and the enforcement mechanisms behind them.

For the downstream relationship once a series is delivered and moving into distribution, working with platforms on contracts, deliverables and timelines covers the three common deal structures and the points where the relationship typically breaks down.