VFX Feedback Loops: Why They Cost More Than Rendering

The most expensive thing in VFX post-production isn't rendering. It is undefined feedback.

That is the observation I would draw after many projects as a producer — and it often surprises people. When productions calculate post costs, "feedback" doesn't appear anywhere as its own line item. There are line items for rendering, for artist hours, for software licences, for storage. There is no line item for "correction rounds that run without a clear note" — even though in most projects that is the single biggest cost.

What actually happens? A shot is delivered, in a first version. Feedback comes in. The artist iterates. Feedback comes in again. Iterates again. A third round of feedback. And so on — until the frame is eventually signed off or until the budget runs out. What actually changes across these rounds is often hard to reconstruct. Sometimes the shot ends up after round five looking surprisingly close to a version that already existed at round one. Nobody says this out loud.

My point in this article: these loops aren't a creative process. They are a schedule and cost risk dressed up as creative process. And they are avoidable — not completely, but to a substantial degree.

"It Doesn't Quite Work"

The first case, which I encounter in almost every project, starts with an email or a Slack message containing exactly that phrase: "It doesn't quite work." Sometimes as a variant: "Something is missing." "Not final yet." "Needs another pass." All four sentences have the same characteristic. They describe an impression without naming an axis.

What does the artist do with this note? In nine cases out of ten, he tries to interpret it — based on what he considers the most likely problem. He makes assumptions. Maybe it's the lighting. Maybe the integration into the plate. Maybe the shadow. He iterates in one direction. The next note arrives: "Hm, now it's even further off." Or, more frequently: "That's it, but something is still off."

What is happening here isn't a creative process. It is a trial-and-error game in which the reviewer holds the key to the solution but doesn't release it — usually because he can't put it into words himself.

My job as a producer in this moment isn't to criticise the reviewer as unspecific. It is to translate the note into an answerable question before it goes to the artist. I send a question back: "Is it the colour temperature, the contrast range, the integration into the plate, the motion path of the element, the sharpness relative to the background?" Five concrete axes. The reviewer picks one or two — or names a sixth that I haven't offered.

This callback takes five minutes. It is uncomfortable for the reviewer, because it forces him to translate a vague impression into a specific observation. But on the other side it saves one or two iteration rounds — and each iteration round costs, depending on the complexity of the shot, between half a day and two full working days.

What I think is important to say here: the reviewer isn't failing when he can't immediately translate an impression into axes. That isn't his work. His work is to assess the frame in the context of the campaign, the brand, the script. The translation into technical axes is my work. It stands between the reviewer's impression and the artist's iteration — and that is exactly where it gets decided how many rounds a shot runs through before it sits.

A sentence I often use in this translation work: "Can we compare this with a reference?" A reference is the fastest way to concretise a vague impression. It forces the reviewer to mirror his "it doesn't work" against a specific other frame — and the frame in which that happens is the fastest way to localise the problem.

Three Reviewers, Three Directions, One Email

The second case is more common than the first — and more expensive. It looks like this: a feedback email reaches the VFX team. In it, in a single block, are notes from three different people. The agency's creative director would like more contrast. The client wants it to feel less dramatic. The director finds the motion of the digital element too fast.

All three notes are legitimate. All three concern the same shot. None of the three notes works alongside the others without prioritisation — if contrast is raised, it becomes more dramatic, not less. If the motion is slowed down, the perceived dynamic of the contrast also shifts.

What happens in the standard case: the artist receives the email and tries to address all three notes. The result is an iteration that doesn't really land in any of the three directions, because it tries to follow three vectors at once. The next review round will inevitably generate new notes, because each of the three reviewers will see in the new version that his vector wasn't implemented.

Three reviewers, three iterations. Instead of one.

My role as a producer at this point is consolidation before forwarding. Concretely: when an email with competing notes comes in, it doesn't go directly to the artist. It goes back to the reviewers — with a short note from me that names the conflicts and asks for a decision. "The contrast note and the less-dramatic note point in opposite directions. Which axis should take priority?" Or: "The motion note and the contrast note influence each other. Please decide which of the two should be implemented first, so we can see the effect in a single iteration."

That isn't pleasant. Reviewers become impatient when they realise their feedback hasn't been forwarded yet. But the alternative is three iteration rounds instead of one. The ratio is worth it.

What I have learned from these consolidations: conflicts between reviewers rarely arise from malice or incompetence. They arise because the reviewers have different roles — brand management, creative, direction, client — and in their respective roles set legitimate but different priorities. It isn't their job to clarify among themselves which priority dominates. It is the job of project management to bring about that clarification, before it lands as an unresolvable knot in the VFX pipeline.

A consolidated note with a clear axis named is gold for the artist. A parallel note with three contradictory axes is an invitation to three iterations — and to costs nobody has budgeted for.

The Iteration That Ends Up Back at Version One

The third pattern is the one in which the futility of the loop becomes most obvious — and which for exactly that reason is most rarely named. It looks like this: a shot runs through several iteration rounds. In round 2, it is adjusted following note A. In round 3, the change from round 2 is partially walked back following note B. In round 4, a note C arrives whose effect ends up looking very similar to what was already there in round 1. In round 5, the shot is signed off in a form that is surprisingly close to the first version.

Nobody says this out loud. It would also be uncomfortable, because it would mean that rounds 2 to 4 generated effort without moving the matter forward.

What ends up visible is a signed-off version — which looks like a normal creative process, but underneath consists of several loops that have cancelled each other out. On the invoice, these loops look like regularly budgeted iteration rounds. In reality, they were costs without yield.

My approach when I recognise such patterns early is to make them visible — and not in the final review, but in the moment they arise. Concretely: when in round 3 a note arrives that walks back what was changed in round 2, I assemble a short version comparison for the reviewers. Three frames side by side: round 1, round 2, round 3 in the proposed direction. With a brief note from me: "The change from round 2 is largely walked back by the current note. Please check whether this is intended, or whether we should return to round 1 or round 2."

That isn't a pleasant step. It forces the reviewers to look at their own process. But it prevents the loop from running for two more rounds before closing at a point that is practically the starting point.

What I want to highlight here: reviewers change their minds — that is normal and legitimate. Assessing a frame isn't a single act; the longer one looks at it, the more often one sees it alongside other frames, the more perception shifts. That isn't a mistake. What becomes a problem is when these shifts aren't reflected — when they manifest as parallel, self-cancelling note tracks in an iteration sequence, without anyone naming the pattern.

Offering the version comparison is my method of making the pattern visible before it becomes a cost.

What Can Be Changed About Correction Loops

Three patterns, one common characteristic. None of them arises from incompetence on the reviewers' side, from lack of care on the artists' side, or from technical limitations of the pipeline. They arise in the gap in between — in the translation from impression to note, from note to iteration, from iteration to next note. Exactly this gap is the place where the most expensive hours of a VFX project are generated.

My point for producers, agency project leads, clients who calculate iteration budgets: the cost of this gap is underestimated in most projects. It is often larger than the cost of the actual creative decisions, and it grows exponentially with every round that runs without a clear axis.

What can be changed about it rarely has to do with technology. It has to do with process. Concretely: with someone who stands between reviewer impression and artist iteration and performs the translation. Who plays unclear notes back before they are iterated. Who consolidates competing notes before they run in parallel. Who makes self-cancelling iterations visible before they become routine.



That isn't spectacular. It is project management in its most invisible form. But it is the lever with the greatest impact on the actual costs of a VFX post — and it sits where nobody sees it on the invoice: not between software and storage, but between person and person.