The revision boundary

Judges limit revisions because a revision fixes a mismatch between what was agreed and what arrived; it does not turn a delivered clip into a first draft of a different one. Without that line, every quote would have to be priced for unlimited redos.

What actually counts as a revision

A legitimate revision request points at something the accepted brief specified and the delivery did not honour - the register drifted from what was agreed, a specific detail the buyer asked to be addressed was missed entirely, or the clip ran meaningfully shorter than the length that was quoted. These are, in effect, quality control on the original agreement: the judge committed to a specific job, and a revision checks that the job matches what was committed to, not a new one. Framed this way, a revision is not a favour a judge extends - it is the tail end of delivering what was already paid for. Payment networks draw the same line: among the card-network reason codes Stripe lists under its "product unacceptable" dispute category, one reads simply "goods or services differ from what was agreed upon for the Payment."

What is not a revision

Wanting a different register than the one you asked for is a new commission, priced as one. Deciding after watching the clip that you actually wanted it longer, or wanted a detail addressed that was never in the original brief, is the same thing: the job changed, so the price changes with it. This distinction matters because the brief and the quote are the recorded agreement between two parties, and a revision request that is really a new brief in disguise is asking the judge to redo paid work for free rather than to correct an actual miss. A judge who cannot tell the difference, or who is pressured into not drawing it, ends up doing two commissions' worth of work for one commission's price.

Why unlimited revisions would break the pricing model

A quote is priced against a specific, bounded job - a given length, a given register, a given amount of judgement time. If "revision" quietly expanded to cover any dissatisfaction, however caused, every quote would need to be priced high enough to cover an unknown number of free redos, which pushes the honest quote for a straightforward request up to cover the cost of buyers who over-use the revision channel. That is the same economic logic that rules out a flat menu price in the first place: a service that has to price for the worst case ends up expensive for everyone doing the ordinary case. A limited, clearly-scoped revision policy keeps quotes honest for the buyers whose requests are delivered correctly the first time, which is most of them.

What a fair policy looks like from the judge's side

Most judges offer one revision, scoped to a genuine miss, within a reasonable window after delivery, and are explicit about that before the quote is accepted rather than after a dispute starts. Being clear about the boundary upfront is itself part of professionalism - a buyer who knows exactly what a revision covers is less likely to ask for the wrong thing and be told no, which is a worse experience for everyone than simply knowing in advance. Stating the policy in the same message as the quote, rather than only after a dispute arises, is what separates a judge who has thought this through from one improvising a boundary in the moment, and buyers can reasonably ask to see it before accepting. A well-run platform holds the brief and the delivery next to each other so a revision claim is a comparison against a record, not a dispute over memory.

The same principle - fix the miss, do not redo the job for free - applies wherever bespoke work is quoted against a stated scope, and it is worth recognising outside judging too: a measurement taken to a stated method either matches the method or it does not, with no ambiguity to revise around, and an automated score has no revision concept at all because there was never a bespoke brief to miss. An AI system sidesteps the whole question the same way, since there is no judgement call in its process to have gone wrong in the sense a revision would fix.