VFX Producer

What they do

The VFX producer turns the supervisor's method decisions into money, people and dates. It starts with a bid: a breakdown where every shot carries a complexity band, a vendor, an estimate and an assumption, because the assumptions are what get renegotiated later. They send packages to vendors, compare bids that are not directly comparable, negotiate awards by sequence, and write the schedule backwards from the delivery date through vendor milestones, turnover dates and review cycles. Once work is running the job is tracking and change control. Every shot has a state, a version, a vendor, a due date and a cost to date, held in a tracking system rather than anyone's head. When editorial changes a cut, the producer establishes what that change costs in reawarded work and lost versions, and takes it to the production with numbers rather than opinions. They manage plate turnovers with the assistant editor, keep the shot count honest as it grows, run vendor reviews and cash flow, and handle the practical apparatus around all of it: crew, visas and remote setups on the production side, invoices and change orders on the vendor side. The unglamorous centre of the job is a spreadsheet or database that has to be right, because everyone downstream plans from it.

What they decide

The calls that belong to this role and no other.

• The bid: complexity banding, assumptions, and what is excluded
• Which vendors are approached and which sequence each is awarded (with the supervisor on capability)
• The visual effects schedule: turnover dates, milestones and review cadence
• What a requested change costs, and presenting that cost before the change is approved
• How contingency is held and when it is released
• Whether a shot is abandoned, simplified or reawarded when it stops converging

Principles of the craft

• A bid is a set of assumptions with numbers attached. The number is not the deliverable; the assumption list is. Most overruns are not mispriced shots, they are shots whose assumptions changed and nobody logged it.
• Shot count is the wrong metric on its own. Two hundred cleanup shots and twenty creature shots are different films with the same count. Band by complexity, and track the bands separately, because the bands fail differently.
• Change is normal; unpriced change is the problem. The producer's job is not to prevent the cut from moving, it is to make sure everyone sees what the move costs before it is approved.
• Turnover discipline is cost control. Every shot turned over before picture is stable will be built at least twice, so a week of patience upstream is often cheaper than any negotiation downstream.
• The tracker is the single source of truth or it is worthless. If any department maintains a parallel private list, the two will diverge and the schedule will be built on the wrong one.
• Protect the vendor's ability to plan. Vendors absorb a certain amount of chaos silently and then absorb it as quality. Predictable turnovers and clear notes buy better shots than pressure does.

Common mistakes

• Bidding to the budget rather than to the work, which moves the overrun later without removing it
• Accepting shots into the count informally, so the total grows without any decision being recorded
• Comparing vendor bids on total price when the bids assume different plate counts, iterations or handles
• Letting the schedule be driven by the last delivery date rather than by the review cycles inside it
• The experienced producer's mistake: managing the tracker rather than the room, so problems are visible in the data a week after the artists knew
• No contingency, or contingency spent in the first third on the shots that were simply first

How to learn it

Exercises and routes in, not a reading list.

• Break down a five-minute sequence from any film into a shot list, band each shot into three complexity levels, and write the assumptions for each. Then have someone challenge every assumption. That conversation is the bid.
• Take a short you have finished and build a full tracker for its effects work with states, versions, due dates and vendors, even if the vendor is you. Run it for two weeks and note where reality and the tracker diverged.
• Price the same shot as a two-week and a five-day job and write what is actually removed at the shorter duration. Being specific about what you lose is the skill.
• Simulate a change: move a cut point in a finished sequence, then list every shot affected, which versions are dead, and what has to be re-turned over. Put a number on it.
• Read a published vendor breakdown of a sequence, then write the bid you think produced it, including the assumptions that were probably wrong.