---
title: "How Much Does an AI Short Film Cost? The $102,000 Breakdown"
url: "https://goremy.ai/blog/articles/ai-short-film-cost-102000-nexus-budget-breakdown"
author: Luis Chavez-Mattos
editor: Luis Chavez-Mattos
editor_title: Director, Product
published: "2026-11-19T18:00:18.242Z"
updated: "2026-11-19T18:00:18.242Z"
topic: AI Image, Video & Audio
description: "PJ Ace says a Nexus episode cost $102,000 and 45 days. Here is the per-person math, what it leaves out, and how to price AI video work."
---

# How Much Does an AI Short Film Cost? The $102,000 Breakdown

> PJ Ace says a Nexus episode cost $102,000 and 45 days. Here is the per-person math, what it leaves out, and how to price AI video work.

Edited by Luis Chavez-Mattos, Director, Product · Published 2026-11-19

## How much does a professional AI short film cost?

It depends on what you count. AI creator PJ Ace said his latest Nexus episode cost $102,000 and took 45 days. By his own split, about 80% of that went to labor and 20% to animation costs. The raw generation cost of AI video is small next to the cost of the people who direct, edit and fix it. The $102,000 figure also comes with caveats. It may cover more than the roughly four-minute episode it was attached to, and it is unclear whether it includes generation credits.

## TL;DR

- PJ Ace's reported **$102,000 and 45 days** covers a roughly four-minute episode, but commentary says it may also cover three or four later parts.
- About **80% of the budget was labor**, which works out to roughly $82,000 across seven people.
- Split evenly, that is about **$11,740 per person**, or roughly **$365 per working day**, which is reasonable and possibly low for skilled video professionals.
- It is **not clear whether generation credits** on the Dreamina platform are inside the $102,000, and they could add up fast.
- A comparison figure of **$1,000 to $1,500 in compute per finished minute** for Junkyard King shows that generation costs can be large in their own right, though the spend is of a different kind.
- If you sell AI video work, **price the labor and the retakes**, not the per-second cost of the model.

## What did PJ Ace actually say about the Nexus budget?

PJ Ace released the latest chapter of his AI series, Nexus, and said the episode cost $102,000 and took 45 days to make. His post reached about 5.7 million views, and it set off an argument about whether that number was reasonable, inflated, or a marketing move.

Three details from the discussion matter before you quote the number anywhere:

1. **The scope is fuzzy.** The post read as if $102,000 paid for one episode of about four minutes. Commentary around it said three or four more parts are coming and the budget covers them too.
2. **Seven people worked on it**, including PJ.
3. **Generation credits may or may not be included.** Dreamina is the platform used for generating video, and it is not clear whether any of the listed allocation covers those credits. If they are not included, the true cost is higher.

Treat $102,000 as a reported figure with an unclear boundary, not an audited production cost.

## How does the $102,000 break down per person?

Take PJ's own split at face value: 80% labor, 20% animation cost. The labor share comes to about $82,000 (the commentary used $82,180). Divided equally among seven people, that is about $11,740 each for the whole production.

Here is how that turns into a day rate:

| Step | Result |
|---|---|
| Labor budget | about $82,000 |
| Split across 7 people | about $11,740 each |
| Spread over 45 days (about 6.4 weeks) | about $1,826 per person per week |
| Assuming a five-day week | about $365 per person per day |

Two caveats. "Labor" is not clearly defined. It could include the director, editors, artists, sound and others, and an equal split is a simplification because a lead earns more than a junior. And a five-day week is generous. Real productions rarely keep one.

## Is $365 a day a fair rate for AI video work?

It is within reason for a skilled video professional, and it may be on the low side. The commentary on this release pointed to School of Motion, which cites day rates of $800 to $1,000 for professional motion work. If those numbers are right, a team earning $365 a day is working below typical rates for traditional production.

That changes how to read the headline. The reaction to $102,000 was often "I could do it for much less." Probably true, especially for someone working alone and counting their own time as free. But the figure is not evidence that AI video is expensive per second of footage. It is evidence that a seven-person team, paid roughly normal wages for six weeks, costs about this much. The same team on a live-action shoot would also be paying for cameras, locations, crew and gear.

## How much does the generation itself cost?

Much less than the labor, per clip. The same source discussion on Seedance 2.5 Flash, a faster, cheaper version of ByteDance's video model, put it at around 15 cents per second of generated video, depending on where you use it. It is about 34% cheaper than the flagship Seedance 2.5, and it only outputs at 480p or 720p. The flagship is described as the most expensive video model in the world, and also as the best.

Here is what that means in arithmetic. A four-minute episode is 240 seconds. At 15 cents a second, one pass over every second costs about $36. Nobody ships a one-pass film, though. The reviewer found the cheap model a "dice roll" that wastes credits on morphing characters, vanishing objects and scenes that fall apart. If you need ten or twenty attempts for each usable second, the bill reaches the hundreds or thousands of dollars. That multiplier, not the sticker price, is the real cost driver. (The twenty-attempt figure is an illustration, not a reported number.)

## What does Junkyard King tell us about compute costs?

Junkyard King is another AI-made series, and the comparison figure cited in the discussion is $1,000 to $1,500 in compute per finished minute. That number is reported, not something verified in the material here, so treat it as a rough benchmark.

Taken at face value, it is a different shape of project from Nexus. For a four-minute piece, $1,000 to $1,500 per minute implies $4,000 to $6,000 in compute alone. That is a small fraction of $102,000. It tells you something useful in both directions:

- **Compute is not nothing.** On a long project it runs to thousands of dollars, and expensive flagship models push it higher.
- **Compute is not the main cost.** Even at the high end, it is a fraction of what seven people cost over 45 days.

The gap between a solo creator spending a few thousand dollars on compute and a team spending $102,000 is mostly the gap between one person's unpaid time and a paid crew.

## How should you price AI video work?

If you take on AI video projects, the Nexus numbers suggest a few rules.

- **Quote labor first.** Use a normal day rate for the role. Don't discount because "the AI does it." The skill is directing, selecting, fixing and editing.
- **Add a retake allowance for generation.** Estimate the seconds you need, multiply by the model's per-second price, then multiply by a realistic retry factor. Show the client the assumption.
- **State what is included.** The Nexus confusion came from an unclear scope. Say whether the price covers one deliverable or a series, and whether credits are inside or on top.
- **Charge for references and prep.** The review found that cheaper models behave better with strong references, storyboards and depth maps. That prep is paid labor.
- **Keep a cost log.** Track hours and credits per project so your next quote is based on your own numbers.

A simple quote template:

```
Project: [name], [length in seconds]
Labor: [people] x [days] x [day rate]
Generation: [seconds needed] x [price per second] x [retry factor]
Tools/subscriptions: [amount]
Contingency: [10-20%]
Total: [sum]
Assumptions: [scope, revisions, who pays for credits]
```

## What does this mean for you?

If you are a **freelancer or small studio**, the lesson is to stop anchoring on the model's price. Clients who see "15 cents a second" will assume the whole job should cost almost nothing. Your quote has to explain that the cost is the judgment and the rework.

If you are a **marketing manager or in-house buyer**, expect AI video quotes to look more like a normal production budget than a software subscription. A quote of tens of thousands of dollars for a polished short is not automatically padded. Ask what is included, how many revision rounds there are, and who owns the generation credits.

If you are a **curious employee** wondering whether AI makes video free, the answer is no. It makes some steps cheaper and others faster, but the human work stays the largest line.

## Could you build your own AI project cost tracker?

If you take on AI video work, the quote template above is exactly the kind of thing that ends up in a spreadsheet nobody maintains. With [Remy](https://goremy.ai), you describe the tool in plain English, such as a project tracker with labor days, day rates, generation credits, a retry factor and a total that updates as you log work, and it builds and deploys a full-stack app you can open from a link. You could add client sign-in, a quote page you can share, and a view comparing quoted costs against actual spend. The app is yours to keep, in a git repo you own.

## Frequently Asked Questions

### Did the $102,000 pay for a single four-minute episode?

Not clearly. The post suggested it did, but commentary around the release said the budget also covers three or four later parts. Without a confirmed scope, you can't calculate an exact cost per minute from it.

### How much of the Nexus budget went to labor?

PJ Ace reported about 80% labor and 20% animation cost. Applied to the reported total, that is roughly $82,000 for labor across seven people, or about $11,740 each. How "labor" was defined is not spelled out.

### Is Seedance 2.5 Flash cheap enough to replace the flagship model?

Not yet, according to the review that tested it. It runs about 34% cheaper and at around 15 cents per second, but it only outputs 480p or 720p and produced more morphing and vanishing objects in several tests. It worked better with many strong references. The reviewer noted it was API launch day, so it may improve.

### Why is the retry factor so important in AI video pricing?

Because per-second prices only apply to successful footage if every generation works. In practice many clips have errors, so you pay for the failed attempts too. A cheaper model that fails more often can end up costing as much as an expensive one that works.
