AI’s Impact on Video Cost Example: An Estimate of What It Should Take to Make an AI Fan Film (And Why the IP Question Should Worry You)
The Video in Question
I found this video and thought it was pretty good. “Darth Vader Arrives on Tatooine… Will He Find Luke? | Part 3” runs eight minutes and twenty-seven seconds. It’s the third installment of a serialized, non-commercial Star Wars “what if” story from a channel called Kyberis — about 3,200 subscribers, 18 videos total, launched in April 2026. YouTube tags it with a “Made with AI” disclosure, and the creator’s own channel page tells you the method in plain language: advanced AI tools build the visuals, while most of the sound — lightsaber swings, blaster shots, cinematic hits — gets crafted by hand I beleive. Every video, the creator says, takes days or even weeks.
Notice what’s missing: no animator credited, no composer, no voice actor, no editor. This is one person doing every job, not a studio with a payroll. Every number that follows — cost, time, skill required — has to be built around a solo operator, not around what a professional animation house would charge to do the same thing.
One Person, Every Job
Think about how many distinct crafts (and tools) get folded into a single eight-and-a-half-minute video.
First, assume everyone has a decent computer and internet connection.
Someone has to write the story, even if it’s a fan sequel riding on someone else’s characters. They have to build the characters so they look the same from shot to shot.
There has to be someone has to generate and curate the actual moving footage, which eats the most hours of anyone’s day. Don’t sleep on lighting or effects.
Then there’s the sound. Someone has to produce a voice track, most likely synthetic since no actor gets billing. There has to be a person to pick or generate music. Someone has to build sound effects by hand, which this creator specifically calls out as their own manual work. Someone has to line all of that audio up against the picture. Finally, someone has to cut the whole thing together, write a title, make a thumbnail, and hit publish.
Every one of those jobs would have its own line item and its own department at a real studio. Here, it’s one laptop and one very patient person.
Where the Character Work Fits — and Why It’s the Foundation
If you remember one technical idea from this essay, make it this one. Before a single second of video gets generated, someone has to lock down what the characters look like — and that step decides whether the rest of the project holds together or falls apart.
Since these are Vader, Leia, and Luke, nobody’s inventing a new look. The job is reproducing an existing design consistently. That starts with generating a small set of reference images — a front-on shot, a three-quarter angle, a profile, sometimes a full body pose — using an AI image tool. That reference set, paired with a written description repeated word-for-word every time, becomes the anchor for every later shot. From there, each new clip gets generated by feeding the video-generation tool that same locked reference plus instructions for what’s happening in this particular scene. Then someone has to check every finished clip against the reference and toss out, and redo, anything that drifted.
If that sounds familiar, it should. This is the AI-era version of something animation studios have done forever: the turnaround sheet and the range-of-motion sheet, drawn once per character so any animator on a team can draw that character correctly from any angle, in any pose, for the rest of the project. The AI workflow swaps a locked set of reference images for the paper turnaround sheet, and a video model’s reference-matching feature for the animator’s hand and eye.
Here’s the catch: a trained animator working from a good turnaround sheet gets the character right basically every time. That’s the entire point of building the sheet. AI reference-matching isn’t there yet. Even when people follow all the best practices, industry guides report only about ninety percent consistency across a ten-shot sequence, and working creators plan on having to throw out and redo fifteen to twenty-five percent of their generated clips because the character drifted (got a different face, a different jaw, wrong hair) somewhere along the way. Close-up shots are the worst offenders, since any small inconsistency in a face reads immediately to a viewer.
AI can attempt the animator’s job of keeping a character consistent through a full range of motion. But it does it with real failure built in, not as a guaranteed skill. That failure rate is exactly why this one step eats a disproportionate share of both time and money, especially for someone who hasn’t already solved it once and built a character library to reuse. This creator clearly has by “borrowing” the same three characters across three installments (so far).
The Part Everyone Skips: Whose Characters Are These, Anyway?
This part is important. A lot of hobbyist (and even some professional) creators wave away and it deserves real attention.
Darth Vader, Luke Skywalker, and Leia Organa are not public domain. They belong to Lucasfilm and Disney. Copyright law gives the owner of a character the exclusive right to make derivative works: new stories, images, or films built on top of that character. A fan film starring Vader, using his name, look, and voice, is as textbook a derivative work as there is.
The video’s own disclaimer admits as much. It states plainly that Star Wars and its characters belong to Lucasfilm and Disney. It claims the video isn’t affiliated with or endorsed by either one. That disclaimer is honest, but it isn’t a legal shield.
Courts have looked at exactly this fact pattern before. A fan-made Star Wars film using the name, the characters, the lightsabers, and the Force. It concluded it looks far more like an unauthorized derivative work than a protected transformation. Putting a disclaimer in the description tells your audience the truth about who owns what. It does not tell a court, or Disney’s legal department, that you had permission.
Layer AI into that picture and the legal exposure gets murkier, not clearer. Regulators and courts here and abroad are actively working through who’s on the hook when an AI tool generates something that resembles a copyrighted character: the person who typed the prompt, or the company that built and trained the model.
The clearest point of consensus so far, from U.S. Copyright Office guidance to recent court rulings, is that AI-generated output doesn’t get its own copyright protection unless a human meaningfully shaped the creative expression.
Separately, and more importantly for our purposes, generating something that closely resembles an existing copyrighted character can still infringe that character’s copyright regardless of whether a human or a machine did the drawing. In other words, “the AI made it, not me” is not a defense that has held up. If your output looks enough like Darth Vader, it’s Darth Vader’s copyright you’re dealing with, whether you drew him by hand or typed a prompt.
Practically, what does that risk look like for a small channel like this one? Disney doesn’t typically sue three-thousand-subscriber fan channels, because the cost of litigation dwarfs any damages. What actually happens is quieter and more common: a copyright claim gets filed against the video. The platform either pulls it down, blocks it in certain countries, or reroutes all the ad revenue to the rights holder instead of the creator. That last outcome is the sneaky one.
The video stays up, everybody keeps watching it, and the person who spent forty, eighty, or a hundred and fifty hours making it earns nothing from it, because YouTube’s copyright system handed the money to Disney automatically. If you’re building a channel, a brand, or any kind of income around content like this, that’s not a hypothetical footnote. It’s the most likely actual outcome. It should factor into whether “borrowing” a famous character is worth it compared to building something you actually own.
How Much Time Does This Really Take?
The creator’s own words — “days or even weeks” — combined with how much trial and error AI video generation still requires, point to somewhere around forty to eighty hours of real work for an experienced solo creator, spread across one to three weeks. That tracks with how often this channel actually posts, roughly once every one to two weeks.
But notice the word “experienced.” That number describes someone eighteen videos into this, with a settled pipeline: characters already locked, prompts already refined, an editing rhythm already worked out. A total beginner does not get to skip any of the hard parts of that learning curve. Just figuring out the tools and their credit systems can burn five to ten hours before any real work starts.
Character consistency, as we just covered, is the single hardest problem in this whole process, and a beginner without an existing reference library will lose far more time and far more money to it than someone who’s already solved it. Getting an AI voice to sound natural and hit the right emotional beats takes real iteration. And hand-building sound effects, which this creator does deliberately, has its own learning curve if you’ve never touched an audio editor before.
A first-timer attempting something of this length and polish should expect something more like eighty to a hundred fifty hours, spread across three to six weeks. (The first attempt or two often doesn’t turn out good enough to publish at all). The forty-to-eighty-hour, one-to-three-week number only becomes realistic once you’ve already made several of these and built the pipeline that makes it fast. This creator is already at that stage. A beginner is a beginner.
What Does This Actually Cost?
For an experienced creator working efficiently, the compute cost — the money spent generating AI video — is the real expense, since everything is metered by the length and quality of footage generated. To land eight and a half minutes of usable final footage, creators typically have to generate somewhere between three and ten times that much raw material, once you count failed takes and alternate versions — so call it fifteen hundred (1,500) to five thousand (5,000) seconds of generated clips.
At current professional-tier pricing, that would run five hundred to three thousand dollars if paid entirely on a per-clip basis, though a flat monthly subscription spreads that cost across several videos a month rather than charging it all to one. All told, an efficient creator’s realistic out-of-pocket spend is somewhere around one hundred to three hundred dollars per video.
A beginner, not being conservative about it, should expect meaningfully more. A first-timer hasn’t learned to prompt efficiently and hasn’t built a character library yet, so figure five to ten times the final runtime in raw generation attempts — for a video this length, that could mean five hundred to a thousand individual tries before reaching a finished cut.
Running through the actual expenses: video generation alone, on a mid-tier monthly plan, will likely require an upgrade or a credit top-up partway through, landing somewhere between one hundred fifty and four hundred dollars for the month it takes to finish one video. Voice generation for an eight-and-a-half-minute script, which runs roughly seven to eight thousand characters of narration, needs at minimum a commercial-use tier and several retries to get the delivery right — call that fifteen to twenty-five dollars.
Character and image generation, especially with a beginner burning extra attempts trying to get well-known Star Wars faces to look right, runs another thirty to sixty dollars. Music and sound effect libraries add another fifteen to thirty dollars. And editing software, if the beginner doesn’t already use a free tool, adds twenty to twenty-five dollars a month.
Add it up and a beginner’s honest, non-conservative first video costs somewhere between two hundred fifty ($250) and five hundred fifty dollars ($500) out of pocket — almost entirely driven by wasted AI video-generation credits rather than by software subscriptions themselves.
What This Would Cost the Old-Fashioned Way
It’s worth pausing on the comparison, because it’s the whole reason this is interesting. An eight-and-a-half-minute animated short with an original score, professional voice acting, and dedicated sound design, made the traditional way through a small studio, would typically run fifteen thousand ($15,000) to sixty thousand dollars ($60,000) or more, and would require a team, including writers, animators, a composer, voice talent, a sound designer, an editor. And it all has to be coordinated.
What this Vader video creator has done is compress that entire team down to one person and a few hundred dollars in subscriptions. That compression is genuinely remarkable. It’s also exactly why the legal question above matters more, not less: when the cost of production drops this far, the volume of content built on someone else’s characters goes up, and so does the aggregate exposure for platforms, creators, and, eventually, the rights holders trying to police all of it.
The Bottom Line
This is a one-person, AI-assisted production, not a studio project, and every number in this essay flows from that fact. Character creation and locking is the foundational step, functioning as the modern replacement for the animator’s turnaround sheet — useful, increasingly capable, but still not as reliable as the trained hand it’s replacing.
The dominant real cost is AI video-generation compute. Experienced creators naturally reduce the amount of spend on the compute by making fewer wasteful (unusable) mistakes. But with compute rates inexpensive at the moment
The dominant resource is the creator’s own time: forty (40) to eighty (80) hours across one (1) to three (3) weeks for someone who’s already built the pipeline, versus eighty (80) to a hundred fifty-plus (150+) hours across three (3) to six (6) weeks for a beginner encountering character consistency, prompting, and AI audio for the first time. Assign an hourly rate to your own time on that.
Many veer towards delegation and that’s wehre paying for others time starts to put porjects into the $1,000’s of dollars.
And sitting underneath all of it is a question worth asking before you start, not after you publish: if the characters aren’t yours, how much of what you build is actually yours? A disclaimer in the description is honest. It isn’t ownership, and it isn’t permission. It may be proof of concept, but don’t be surprised if you are forced to take it down.
Sources: YouTube video and channel data; Kling AI pricing; Kling AI pricing (PhotonPay); ElevenLabs pricing; AI character consistency guide (Pixo); Character consistency across shots (Pixmind); Kling 3.0 character consistency guide; Long-form AI video character consistency guide (AI Magicx); Character consistency state of the art 2026; Congressional Research Service on generative AI and copyright; University of Akron law review on fan films and copyright; DLA Piper on AI-generated content and copyright liability
More on the Media Production Business
Sports Media for Entrepreneurs