Billing for direction, not revision rounds
Round counting priced a production cost that no longer exists. Price changes of direction, and make refinement unlimited inside a locked direction.
Three revision rounds was never a creative principle. It was an estimate of how many times you could re-grade, re-cut or re-shoot before the job stopped being profitable. The number tracked a real constraint, so nobody had to justify it. Remove the constraint — make another version of a shot take minutes instead of a day — and the clause survives as a habit while the thing it measured has gone.
You can tell it has gone by the shape of the arguments it now produces. A client sends four precise, easily-actioned notes in one document and burns a round. Another client sends "not quite it" three times in a week and burns three. The second one costs you far more and is charged far less. That is a pricing model measuring the wrong variable.
What actually got expensive
The input that stayed expensive is the decision. Specifically: the decision about what the thing is.
Generation is fast; deciding is not, and deciding badly is what costs real money because it invalidates work downstream. Superside's Breakpoint research found 80% of creative teams at or beyond capacity and 70% of creative leaders burnt out — despite AI adoption, not before it. The pressure moved rather than eased. Volume is up, the bottleneck slid downstream into filtering, brand governance, rights and taste, and none of those get cheaper by making the render faster.
That is the thing worth pricing: not how many times you pressed go, but how many times the brief changed underneath you.
There is a second reason to move off round counting, and it is defensive. Billing by the unit of production effort — hours, rounds, renders — puts your own cost structure on the invoice for a client to negotiate against. If AI turns a twenty-hour deliverable into a five-hour one and you are still billing hours, you have just cut your own invoice by 75% and called it efficiency. Fixed-price versus cost-plus quoting for video walks through that arithmetic properly.
Define a direction, then define what sits inside it
The model only works if "direction" is written down as precisely as a spec. A direction is the set of decisions everything downstream depends on:
- The concept — what happens in the piece, and what it is arguing.
- The look — the approved reference set, the palette, the grade.
- The voice — script, tone, and who or what delivers it.
- The format — duration, aspect ratio, and the platform it is cut for.
Lock those four at the concept gate. Everything that does not change one of them is a refinement, and refinement is unlimited.
| Request | Classification | Why |
|---|---|---|
| "Hold the product shot a beat longer" | Refinement | Timing inside the locked cut |
| "Warmer grade, keep everything else" | Refinement | Look is unchanged, the grade moves within it |
| "Swap the second line of the VO" | Refinement | Script edit, same argument |
| "Try the other music bed" | Refinement | Chosen from the approved options |
| "Can the presenter be older?" | Direction change | New casting invalidates every generated shot |
| "Make it funny instead of aspirational" | Direction change | New concept, new script, new everything |
| "Let's lead with the discount, not the story" | Direction change | New argument |
| "Also give us a 16:9 version" | New deliverable | Separate asset with its own placement |
Two things fall out of that table immediately. Refinements are the majority of client notes by volume and the minority by cost. And direction changes are rare, expensive, and currently invisible in most proposals — they arrive dressed as "just one more thought."
The pricing structure
Three lines replace the round cap.
One: the base fee buys one direction, taken to acceptance. Not one round, not three. One direction, refined as much as it takes to meet the acceptance criteria, inside the agreed feedback window.
Two: a direction change is quoted before it happens, at a stated price. Publish the number in the proposal so it is never a surprise and never a negotiation in the middle of a job.
Three: late direction changes cost more than early ones. This is the part clients find counter-intuitive and accept instantly once it is explained, because it is obviously true: a concept change at the brief stage throws away a conversation, and the same change after key frames and motion are approved throws away everything built on top of them. Price it as a ladder rather than a flat fee — cheapest before anything is generated, most expensive after the assembly is signed off.
There is no public benchmark for what those numbers should be. I looked; a survey, contract corpus or rate study on how studios restructured revision terms once generation became cheap does not exist in the public record. What does exist is the ordinary rate context you can sanity-check a base fee against: per-minute quoting for AI video work is commonly advertised somewhere in the $50–$150 per finished minute range, and freelance retainers commonly start around $1,500 a month and tier up from there. Those are advertised ranges rather than a survey, so treat them as orientation. Set the direction-change fee as a proportion of your own base rather than copying a figure, because a number that is meaningful against one project is noise against the next. Per-minute versus per-deliverable versus retainer pricing covers choosing the base unit, and how to price AI video services as an agency covers the full ladder.
Why unlimited refinement is affordable
Because refinement mostly happens before anything is charged. The editor is one re-renderable timeline: preview: true renders a 480p pass at no credit cost, with a short per-user cooldown, and the export charge lands once on the version you confirm — regardless of how many clips are on the timeline. Iterating on free 480p previews before you pay for the export is the mechanic in detail.
So "warmer, then back, then warmer but keep the crop" is a sequence of preview renders and one export, not three billable versions. That is what makes the promise safe to make. What it does not make free is your attention, which is why the feedback window and the named approver still matter — the constraint you are protecting is calendar, not credits.
Keep the two ledgers separate in your own books, though. Tracking credits per client deliverable is how you find out which accounts are quietly expensive, and that is a margin question you answer internally rather than a line you put on the invoice.
The "shouldn't this be cheaper now?" conversation
Price this model and you will eventually be asked to discount it. The published data suggests you will be asked less often than you fear. Reporting aggregated by TechBullion puts 73% of agencies as never having been asked to cut prices despite adopting AI, and among the 27% who did face a discount request, only 13% actually lowered their rates. Clutch's 2024 figure — 61% of agency clients raising AI during renewals — is about the topic coming up, not about the price coming down.
The answer that works is not a defence of your process. It is a restatement of the unit:
"The fee isn't for the hours, it's for the direction and for getting it right. What's changed is that you can now ask for as many refinements as you want without me billing you for each one — that's where the speed goes. If we change the direction itself, that's quoted, and you'll always know the number before you decide."
Two sentences of value framing beat a defensive itemisation every time, and the second sentence is the one that makes the first believable, because it puts a real constraint on the promise.
If you are migrating an existing account, do it at renewal rather than mid-project, and lead with what improves for them. Raising prices on existing content clients covers the sequencing. Productized clip tiers and where they break is worth reading first if your instinct is to solve this with packages instead.
FAQ
Doesn't unlimited refinement invite abuse?
Occasionally, and the fix is a clock rather than a cap. Refinement is unlimited inside the feedback window and inside the locked direction; notes arriving after the window reopen the project and are priced. In practice the clients who would abuse it are the ones changing direction repeatedly, and those changes are exactly what this model charges for.
How is this different from just capping rounds at three?
A round cap prices the number of times you press go, which is nearly free. This prices the number of times the brief changes, which is expensive. The practical difference shows up with the client who sends careful, thorough notes: under a round cap they subsidise the vague client, and under this model they do not.
What if a client refuses to accept that something is a direction change?
That is an acceptance-criteria problem, not a pricing one. If the reference set and the concept were signed off in writing at the concept gate, the classification is a comparison rather than an opinion. The revision policy that stops scope creep has the clause language for drawing that line before the argument happens.
Should the direction-change fee ever be waived?
On the first project with a new client, plan to waive one. The value of the clause in month one is that it exists and gets mentioned out loud; enforcing it from project two is easier because you both have a shared reference for what a direction change actually cost in time.