How to Budget for Retakes and Revisions on an AI Vertical Drama Commission

H1: How to Budget for Retakes and Revisions on an AI Vertical Drama Commission

Batch four arrives. Nineteen shots need work. Nobody wrote down who pays for them.

That is the moment most first commissions discover they budgeted for delivery and not for correction. The production agreement names a total fee, a delivery date and a set of technical specifications. It says very little about what happens when a shot misses those specifications, how many attempts are included, or at what point a request stops being a fix and becomes a new instruction. The result is a commercial negotiation held at the worst possible moment, mid production, with a delivery window closing and both sides already committed. A retake reserve removes that negotiation from the critical path. It is the least interesting line in a commissioning budget and the one that most reliably decides whether the series lands on the number you approved.

1. The Revision Line Item Most Commissions Leave Blank

Look at a first vertical drama commissioning budget and you will usually find script, generation or shoot, voice, edit, music, localisation and delivery. Revisions sit inside one of those numbers, unnamed and unsized. The producer has assumed some allowance. The commissioner has assumed more. Neither figure exists on paper, so the gap only surfaces when it is consumed.

The reason this happens is that revision volume feels like a quality question rather than a budget question. It is treated as something a good partner simply absorbs. In practice it is a capacity question. Every corrected shot occupies a generation slot, a review slot and an edit slot that were allocated to something else. A production running 900 shots across 70 episodes with a ten percent correction rate is running 990 shots. That is a real ninety shot workload, and it lands somewhere whether or not anyone budgeted for it.

Naming the line changes the conversation from blame to arithmetic. You are no longer asking whether a shot should have been right the first time. You are asking how much correction capacity the series needs, who holds it, and what happens when it runs out. Both sides can plan against that. Neither side can plan against an unnamed assumption.

2. What Actually Counts as a Retake

The single most expensive ambiguity in a commissioning agreement is the boundary between a defect and a change. A defect is output that fails an agreed specification. A change is output that meets the specification and is no longer what you want. The first is the producer obligation. The second is a new instruction with a cost attached. Most disputes are not about money. They are about which of those two categories a note falls into.

Defects that should be corrected at no additional cost

Write these into the agreement as named failure classes rather than a general quality promise. A character whose face, hair or wardrobe does not match the approved reference. A shot that misses the technical delivery spec on resolution, frame rate, aspect ratio or loudness. Dialogue that does not match the locked script. A continuity break against the series bible that was supplied before production. Artefacts, warping or duplication that any reviewer would flag on a first pass. These are objective. A third party could adjudicate them from the documents.

Changes that consume the reserve

A director preference on framing where the original framing matched the brief. A performance read you want warmer after hearing it. A location or wardrobe substitution driven by a note from marketing. A new cliffhanger placement after the release plan shifted. None of these are failures. They are you changing your mind, which is legitimate and often correct, and which costs generation and edit time that was allocated elsewhere. The discipline of scope control is well established outside media, and the framing used in the broader project management field, where change requests are logged and priced rather than quietly absorbed, transfers cleanly onto a production slate.

The practical test is whether a note can be traced to a document that existed before the shot was produced. If it can, it is a defect. If it cannot, it is a change. That single sentence in an agreement prevents more conflict than any quality clause.

3. How to Size the Retake Reserve Before You Sign

Start from shot count rather than percentage of fee. Count the deliverable shots, multiply by an expected correction rate, and price that volume at the same unit economics as the base production. A reserve expressed as a shot allowance is enforceable and legible. A reserve expressed as a vague percentage invites both sides to interpret it favourably.

For a first series with a partner you have not worked with, plan for a correction rate in the range of eight to fifteen percent of total shots. That range narrows quickly. By the third series with the same partner and the same reference set, five to eight percent is realistic, because the character library, the prompt structure and the review standard are all established rather than being discovered during production. Sequels on an existing franchise sit lower still.

Hold the reserve in two parts. The first part is the producer allowance, included in the base fee, covering defects. The second part is your change reserve, held on your side of the budget, covering notes that are not defects. Keeping them separate stops a producer allowance from being quietly consumed by creative changes, which is how a series arrives at episode sixty with no correction capacity left and a genuine defect on the table.

Set a cap and a trigger. The cap says how many corrected shots the base fee covers. The trigger says what happens when the cap is reached: a written notice, a joint review of why the rate is high, and a priced extension rather than an argument. Sizing the reserve properly is downstream of sizing the commission properly in the first place, and the same unit economics apply.

4. Structuring Revision Rounds So They Close

Unlimited revisions do not produce better series. They produce series that never lock. The structure that works is a small number of defined rounds with a closing condition on each, applied per batch rather than to the whole series at once.

Two rounds per batch is the workable default. Round one is the full note pass: every reviewer submits against the batch inside a fixed window. Round two is confirmation only, checking that round one notes were executed, not opening new ground. Anything raised in round two that is not a round one note becomes a change request against the reserve. That rule is what makes the round close.

Batch size matters more than round count. Ten episode batches give you a correction signal early enough to act on and small enough to review properly. Reviewing a whole series at once produces a note list so large that patterns disappear inside it, and by then every systemic error has been repeated sixty times. A ten episode batch that reveals a wardrobe inconsistency in episode three saves the same error in episodes eleven through seventy.

Fix the review window in the agreement, in working days, and make it binding on you as well as on the producer. A commissioner who takes eleven days to return notes on a five day window has consumed production schedule, and every downstream date moves. Delivery dates that slip because of commissioner review latency are the most common schedule failure on a first commission and the easiest to prevent.

5. The Notes Discipline That Keeps Retake Volume Down

Retake volume is driven less by generation quality than by note quality. A note that says the scene feels flat produces a guess. A note that names the shot, states what fails, points at the reference and specifies the intended outcome produces a fix. The second note costs the reviewer ninety seconds more and saves two generation cycles.

Adopt a single note format and refuse notes that do not use it. Shot identifier. Observed problem. Reference or specification breached. Required outcome. Priority. One note per issue rather than a paragraph containing four. Notes that arrive as a voice message or a comment thread on a video player lose the shot identifier immediately, and a note that cannot be tied to a shot cannot be executed or audited.

Consolidate before sending. A single reviewer of record collects notes from creative, brand, legal and localisation, removes contradictions and sends one list. Producers receiving three unreconciled note sets will either pick one or ask which governs, and both outcomes cost a cycle. Reconciliation is a commissioning side job and it belongs to a named person.

Watch for the note that is really a brief problem. If a reviewer keeps rejecting shots that match the approved brief, the brief was wrong and no amount of correction will fix it. Reopen the brief, pay for the change, and stop spending the defect allowance on a specification error. Acceptance criteria defined before work begins are the mechanism that makes this visible, and the principle sits at the heart of the ISO 9001 family of quality management standards, where conformity is measured against a documented requirement rather than against a reviewer mood on the day.

6. What to Watch in the First Two Batches

The first two batches tell you almost everything about the retake budget for the remaining series. Track three numbers and nothing else at this stage.

Correction rate by cause. Group corrected shots into character consistency, technical spec, script mismatch, continuity and preference. A rate of twelve percent that is ninety percent preference is a briefing problem on your side. The same twelve percent driven by character drift is a pipeline problem on the producer side. The headline number alone tells you nothing actionable.

Repeat correction rate. Shots corrected twice are the real cost signal. A first correction is normal. A second correction on the same shot means the note was unclear, the reference was ambiguous or the underlying method cannot hit the target. Any repeat rate above roughly one in ten corrected shots warrants a conversation before batch three, not at delivery.

Rate direction. Batch two should be below batch one. If it is flat or rising, something systemic is unaddressed and the reserve will be gone before the series is. Falling rates across batches are the clearest evidence that a partner is learning the series rather than producing it shot by shot.

7. When a Retake Bill Is a Symptom, Not a Cost

Sometimes the reserve is not the problem. A correction rate that stays above twenty percent through three batches is not a budgeting failure. It is evidence that the series is being produced without a stable reference set, without a quality gate before delivery, or against a brief that was never specific enough to hit.

The remedy is structural rather than financial. Stop the batch. Rebuild the character reference set and lock it. Establish a producer side review gate so that shots failing an objective spec never reach you at all, which is a normal part of a functioning pipeline rather than a favour. Re-approve a short sample batch before resuming volume. That pause costs schedule and saves the reserve, and it converts an escalating cost into a fixed one.

The inverse case matters too. A correction rate near zero in batch one is not automatically good news. It sometimes means your review is too shallow to catch what platform acquisition review will catch later, which is the most expensive place to discover a defect. A reserve that goes unused because nobody looked hard enough is not a saving.

Axis AI Studios Perspective

Axis AI Studios treats the retake reserve as part of the production plan rather than a contingency. Shot counts, correction allowances, review windows and the defect versus change boundary are set before a series enters production, because every one of them is cheaper to agree in advance than to argue about in batch four. The reserve is sized from the shot list, split between defect allowance and commissioner change capacity, and reported against batch by batch so that a rising rate is visible while there is still series left to protect.

We are an AI native vertical drama production studio working with platforms, media companies, brands and IP holders across commissioning, production and delivery. Current production partners include Den Tolmor and Good Fight Production LLC and HolyWater. Recognition of the operational problem comes first. A commissioner who has run a slate knows that correction capacity is where budgets actually fail, and a producer who will not put a number against it has not run one.

If you are sizing a first commission and want the revision structure specified before the fee is agreed rather than after, write to business@axisaistudios.com.

FAQ

What is a reasonable retake allowance to negotiate into a first AI vertical drama commission?

Express it as a shot allowance rather than a percentage of fee. For a first series with a new partner, an allowance covering ten to fifteen percent of total deliverable shots for defects is defensible, with a separate change reserve held on the commissioning side for notes that are not defects. By the third series together the defect allowance should be materially lower, because the reference set and the review standard are established rather than being built during production.

Who pays when a correction is caused by an unclear brief?

The commissioner, in most reasonable agreements, which is exactly why the boundary needs defining before production starts. If a note cannot be traced to a document that existed before the shot was produced, it is a change rather than a defect. The useful consequence of writing that rule down is not that it wins arguments. It is that it makes briefing quality a visible cost, which improves briefs quickly.

Should revision rounds be capped, and what happens when the cap is hit?

Cap them, at two rounds per batch, with round two limited to confirming that round one notes were executed. When the cap is hit, the mechanism should be a written notice and a joint review of the correction causes, followed by a priced extension if the volume is genuinely required. An uncapped revision process does not protect quality. It removes the point at which anyone has to diagnose why the rate is high.

Further Reading

For the base numbers that a retake reserve sits on top of, the first commission budgeting breakdown covers what each production line item actually costs across live action, AI native and hybrid models.

For the contractual language that makes review windows and quality gates enforceable rather than aspirational, the production SLA guide covers the specific provisions worth writing into an agreement.

For the producer side discipline that should catch defects before they ever reach your note list, the quality control method for large shot volumes covers how a review pass is structured across hundreds of generated shots.