Quoting a shot you have never generated

There is a specific shot in most interesting briefs that you cannot price, because you genuinely do not know whether it works. Not "it will be hard" — you have no idea whether the first attempt looks right or whether the fortieth does.

Most operators handle this one of two ways, and both are bad. They pad the quote by a guessed percentage, which prices the risk without resolving it and loses the pitch to someone who guessed lower. Or they quote optimistically, discover on day three that the shot does not land, and absorb the difference silently. The second is far more common, and it is why studios with healthy revenue run thin on projects that looked fine on paper.

The fix is structural, not arithmetic. Do not price the unknown. Price the process of removing it.

Which unknowns are actually unknown

Not every hard shot is unpriceable. Sort them before you quote:

Known-hard. You have made shots like this. Hands on a product, a face in motion, matched continuity across two clips. These have a difficulty tier and an attempt allowance and go straight on the rate card.

Unknown-but-adjacent. Something structurally similar with a different subject — a liquid pour when you have only ever done a powder pour. Risky, but you can reason about it. Quote a tier above your instinct and move on.

Genuinely unknown. No adjacent reference in your own logs: a specific mechanical action, a named real-world location, a culturally specific gesture, an animal doing something particular. Here you are not estimating, you are guessing, and the honest position is that the hit rate is unknown until someone tries.

Only the third category needs a contingency structure. Apply it everywhere and clients stop reading it.

The paid test block

A test block is a small, fixed-fee, separately invoiced phase whose only deliverable is an answer.

Four properties make it work:

1. Fixed fee, not an estimate. Whatever number you pick, it does not move. The client is buying certainty about the cost of finding out, which is the entire product.
2. A defined output that is not the shot. The deliverable is a written finding plus the best attempts as 480p previews, not a finished asset. Say this in the quote in plain language, because a client who thinks they are buying a shot will be disappointed by a memo.
3. A time box. Two working days is usually right. Long enough to try several models and prompt strategies, short enough that it does not become the project.
4. A stated decision point at the end. Three outcomes only: it works at the quoted tier, it works at a higher tier and here is the number, or it does not work and here is the fallback.

It also fixes the sales conversation in a way padding never does. "That shot is genuinely uncertain, and here is a small fixed fee to find out before either of us commits" reads as competence. A quietly padded quote just reads as expensive.

Run test blocks against a deposit rather than net terms. You may be about to tell the client something they do not want to hear, and you want that invoice already settled — deposits, net terms and kill fees covers the shape.

The attempt ceiling

If the client will not fund a test block, or the shot is only one line in a much larger project, put a ceiling in the quote instead.

An attempt ceiling is a number of generations included in the fee for that specific shot. Past it, one of three things happens, and the quote says which:

• The shot moves to the next difficulty tier and is billed at that rate;
• The shot is replaced by the fallback deliverable at no change in fee;
• Work on the shot stops and the fee is reduced by the shot's line price.

Pick one per project and write it. The third option is the one clients trust most, because it is the only one where the risk visibly sits with you as well.

Set the ceiling from your own logs, not from the attempt ratios that circulate in community write-ups. Those are aggregated anecdote, and no measured hit-rate study exists in the public record to check them against. Measuring your own reroll rate is the only version of the number worth putting in a contract.

A useful default: set the ceiling at three times your median attempt count for the nearest comparable shot you have actually made. Wide enough that it rarely triggers, tight enough that it means something when it does.

The fallback deliverable

The clause that makes the whole structure acceptable to a client is the one that guarantees they get something. Without it, the contingency reads as "we might not deliver, and you might pay anyway."

A fallback is a pre-agreed alternative shot that covers the same beat in the edit. It is chosen at quote time, priced at quote time, and needs no new approval to invoke. Common ones:

Failing shot	Workable fallback
Specific hand action on a product	Cut on the action — before-state and after-state as two static-camera shots
Character speaking to camera	Voice-over over b-roll of the same character, no lip sync required
Complex mechanical motion	Animated still with a slow push and a subtle parallax
Real named location	Establishing shot from licensed stock, generated coverage for everything else
Legible pack copy in frame	Clean pack shot, copy composited on the timeline

Every one of these is a real shot a real editor would use. That is the test: if the fallback would embarrass you in the finished cut, it is not a fallback, it is a failure with better branding.

Notice how many of them work by moving the problem out of generation and into the edit. Versely's editor is EDL-based, so composited text, graphics and audio live on a re-renderable timeline and the final export is charged once regardless of clip count — which is exactly why those fallbacks are cheap.

Clause language

Paste this into the scope section and adjust the numbers:

Shot 7 (uncertain feasibility). This shot has no direct precedent in our production history and its feasibility is not established at signature. It carries an attempt ceiling of 24 generations. If an acceptable take is not produced within that ceiling, we will deliver the fallback shot described in Appendix A at no change to the project fee, and Shot 7 will be considered delivered. If the client requires further attempts on the original concept beyond the ceiling, those are billed at the T4 rate in the attached rate card.

Four sentences, and it removes every version of the argument you would otherwise have in week three. Note that it does not promise the shot. A quote that stays silent about feasibility is, functionally, a promise you may not be able to keep. If the brief has several of these, you are not in fixed-price territory at all — you are in base plus change order, and pretending otherwise just relocates the loss.

Running the block so it settles the question

A test block that produces "hmm, maybe" is worse than none, because you have spent the client's money and still have to guess.

Design it to be conclusive:

1. Fix the prompt, vary the model. Same subject, same framing language, dispatched across several named models in one go. The agent chat can fan a single prompt across multiple models in one request, which is the fastest way to find out whether the difficulty is the brief or the routing. Building a prompt suite for testing video models is the structured version.
2. Then fix the model, vary the prompt. Once one model looks least-bad, spend the rest of the block on prompt strategy against that one. Mixing both variables at once tells you nothing.
3. Iterate at preview resolution. The editor's 480p preview pass is free and carries a short per-user cooldown, so it is not an unlimited loop, but it is the right resolution for deciding whether a take is structurally right before committing to a final export.
4. Diagnose, do not just reroll. If every attempt fails the same way, the problem is the prompt or the model, not luck. A diagnostic tree for generations that come back wrong is faster than another twenty rolls.
5. Write the finding down. One paragraph: what you tried, what happened, what tier the shot really is. That paragraph is the deliverable the client paid for, and it is also the log entry that moves this shot out of "genuinely unknown" forever.

That last point compounds. Every block you run permanently shrinks the category of shots you cannot price, and after a year of logging them most briefs contain none at all.

FAQ

What do I charge for a test block?

Enough that a day of your attention and the credits are covered, and low enough to be an obviously reasonable line next to the project fee. It should land well under the price of the shot it is de-risking, which is the argument you make on the call.

What if the client refuses to pay for a test?

Quote the attempt ceiling with the fallback and skip the block. You still get a bounded downside, you just find out later and with less room to change the plan. If they refuse the ceiling too, the project is asking you to underwrite an unknown, and no rate fixes that.

Can I run the test before I send the quote, to win the pitch?

Sometimes, on work you want badly, and treat it as sales spend rather than production. The trap is doing it habitually: unpaid feasibility work on every pitch is a large unbilled workload with a low conversion rate attached.

How does this interact with a fixed project fee?

Cleanly, as long as the fallback sits inside the fee and the ceiling is written. The fee covers the shot or its fallback, whichever ships. What is not covered is unlimited attempts on the original concept, and saying so is what keeps the fee fixed for you too.