
Price AI film work by the cost of a shot the client accepts, not by the credits it takes to render one. This article puts three public figures from November 2026 side by side and shows that each one points the same way: generation is a small, visible line item, and judgement, revision and review are where the money goes.
Method. Sources reviewed November 18, 2026. All figures are as reported by the people and outlets named. We did not audit any of these budgets. Our own arithmetic is labelled.
What this means for you
- Price your work per accepted shot, not per credit or render. In the one episode budget we could break down, sequence runs were roughly 1.3% to 5.1% of the reported total.
- Batch your notes. Ten notes sent together cost one extra render pass; the same ten sent one a day cost ten.
- Lock the previs and agree which version is being judged before you start generating revisions.
- When someone quotes a saving like "70–80% fewer hours", ask which step it came from and what happened to review and comp time afterwards.
- Know which of your shots could move to open models if a paid provider raised prices, and check you're allowed to use them.
Three numbers from November
| Figure | Source | What it measures | What it doesn't |
|---|---|---|---|
| $102,725 and 45 days for one Nexus episode | PJ Ace, on X | Total reported cost and time for a ~15 min episode | How it splits between labour, tools and overhead (beyond "mostly labor") |
| KRW500M (about $366K) in tokens for a 100-min feature | Kim Kwang-jib, Studio Meta-K, at TCCF Taipei | Token bill alone, at premium commercial tools | Labour, review, comp |
| Up to 70–80% fewer work hours | Fuji TV | Savings on specific tested production steps | Savings on a whole show or the whole pipeline (Fuji says so itself) |
Where $102,725 went
PJ Ace's answers further down the thread show where the money went:
- each 30-second sequence is run 3 to 7 times, at roughly $15 to $25 a run
- Claude splits the script into scenes and 30-second prompts, with animators telling it what each shot has to hit
- there was no 3D or Blender on this episode
- the episode runs about 15 minutes across three parts, and he says the $100k+ is mostly labour
Our arithmetic (illustrative, from his stated ranges): 15 minutes is about 30 sequences of 30 seconds. 30 × 3–7 runs × $15–25 comes to between $1,350 and $5,250 in sequence runs, or roughly 1.3% to 5.1% of the reported total. Character sheets and location plates add to that, but generation stays a small slice either way.



