LEARN · COSTS & BUDGETING

What does an AI short film actually cost?

TL;DR
  • A 3-minute AI short film can cost roughly $50–$300 in new tool spend, if you already own an editing program. Most of that money goes to image and video generation credits.
  • My horror short All the Pretty Seahorses (3:14) cost about $100 and took about 4 days, roughly 24 hours of work. I made it in OpenArt (character and world images with Nano Banana and GPT Image 2.5, Seedance video, voice cloning), then used Topaz to upscale, Suno for music, Final Cut Pro to edit and ChatGPT to develop the story.
  • My tech-noir short FEMTECH was made in Higgsfield. It cost about $200 and took 20–30 hours over 4–5 days.
  • The biggest hidden cost is rejects. You pay for every take you throw away, so lock your characters early and draft in low resolution.

Every time I post a clip, somebody asks the same question: "Okay, but what did that cost?" Fair question. Most of the numbers you see online are either a hype figure ("I made a movie for $0!") or vague ("it depends"). Greenlight is a studio I'm building in public, so here are my real numbers from two finished shorts, what the money actually paid for, and how I'd budget the next one.

One caveat before we start. AI tool pricing changes constantly. Plans get renamed, credit costs shift and new models come out every few weeks. So I'm giving you ranges and proportions, not a per-clip price list that'll be wrong next month. Always check the current pricing page before you subscribe.

The short answer: $50–$300 for a 3-minute short

If you already own an editor (Final Cut Pro, DaVinci Resolve's free version, CapCut or Premiere) and you're making a short of around 2–4 minutes, I think a realistic budget is $50 to $300 in new subscriptions and credits. Where you land in that range depends on three things:

  1. How many shots you need. More cuts means more generations.
  2. How picky you are. If you'd reject 4 takes for every one you keep, you're paying for 5 clips to get 1.
  3. Whether you have recurring characters. Keeping the same face across 30 shots takes setup and extra re-rolls.

Here's how that played out on my two films.

Film 1: All the Pretty Seahorses, about $100 and 4 days

All the Pretty Seahorses is a 3-minute Florida coastal horror short. A travel vlogger and his cameraman head down to the Keys to film a reclusive "seahorse whisperer," and learn why you never touch the seahorses. You can watch it on the case study page.

JOBTOOLHOW IT HIT THE BUDGET
Story developmentChatGPTNot a project cost for me (it's a general-purpose subscription)
Characters & world stillsOpenArt image models (Nano Banana, GPT Image 2.5) + character toolsCredits on a mid-tier OpenArt plan
MotionSeedance, inside OpenArtSame credits. This was the biggest share
VoicesMy own performances, cloned with OpenArt's voice toolSame credits
UpscaleTopazRolled into the total
ScoreSuno (paid tier, for commercial rights)Rolled into the total
EditFinal Cut ProSoftware I already had
TotalAbout $100, ~4 days, ~24 hours

I deliberately didn't break that $100 down tool by tool. Honestly, I didn't track it that closely while I was making the film, and a made-up split would be worse than no split. What I can tell you is the shape of the spend. Most of it went to generating video. A lot of that video ended up on the cutting-room floor.

Three choices kept it at around $100:

  • I drafted motion at 480p. Low-res drafts are much cheaper. I only upscaled the takes I was actually keeping, to 2K (2560×1440) with Topaz. Upscaling a reject is pure waste.
  • I locked characters before shooting. I generated several variants of Keegan, Trevor and Nereus, picked one of each and saved it as a character. Every later shot references that locked character. I call this the reject → lock loop. It cuts down the "why does he have a different nose now" re-rolls.
  • I performed the voices myself. I recorded every role on my phone, then cloned my own voice into each character. My performance keeps the timing and comedy human, and I didn't need a separate voice-actor budget.

Film 2: FEMTECH, about $200 and 20–30 hours

FEMTECH, my tech-noir short, is a tech-noir coastal thriller about a former tech engineer whose stolen invention becomes the weapon of her revenge. I made it entirely in Higgsfield, originally for the Higgsfield Global Film Festival. It cost me about $200 and took 20–30 hours over 4–5 days.

So why did it cost roughly double? I'll be straight: this isn't a clean lab comparison. It's a different platform, a different visual style and a different story, and I put in more hours. The pattern I've noticed holds across both films, though. Your spend tracks your generating hours far more than your runtime. Both films are about three minutes long. The one I spent more hours generating cost more. More hours means more takes, and more takes means more credits.

That's the most useful way I know to think about AI film costs. You aren't really paying per finished minute. You're paying for every attempt.

Where the money actually goes

Across both projects, this is roughly how I'd rank the costs, biggest first:

  1. Video generation. By far the largest line. Every clip costs something, and high resolution, longer duration and premium models cost more.
  2. Rejects and re-rolls. Technically part of #1, but worth calling out on its own. A take with a melted hand, a wrong eyeline or a character who changed outfits still costs full price.
  3. Image generation for characters and world stills. It's cheaper per item than video, but it adds up when you're building a cast and a dozen locations.
  4. Voice. Cloning your own voice is cheap. Hiring actors or licensing voices costs more.
  5. Music. A paid Suno tier is a small monthly cost, and it matters (more on that below).
  6. Upscaling and editing. Usually a one-time or already-owned tool. Free editors exist.

Time is the cost nobody quotes

"Made in a weekend" is technically true for a lot of AI shorts, mine included. But 24 hours of focused work on Seahorses and 20–30 hours on FEMTECH is still a real chunk of time. Most of it isn't spent clicking "generate." It's spent judging takes, rewriting prompts, matching continuity and editing. If you value your time, count it. If you're doing this to learn, the hours are the tuition, and they pay off. My second and third films go a lot faster than my first.

How I'd budget your first AI short

These are my ballpark ranges for new spend, assuming you already have an editor. They're estimates from my own experience, not quotes:

PROJECTBALLPARK SPENDBALLPARK TIME
Test piece / teaser under 1 min$0–$40 (free tiers can cover it)A few evenings
2–4 minute short with recurring characters$75–$25020–35 hours
5–10 minute short$200–$600+40–100+ hours

A few rules I follow now:

  • Write the script before you generate anything. Every scene you cut on paper is credits you never spend.
  • Set a cap and track spend as you go. Write it down during production. Afterward you won't remember.
  • Budget around 2× what you think the keepers will cost. Rejects are where the money goes.
  • Draft low-res, then upscale only the keepers.
  • Use one platform per film where you can. Splitting credits across five subscriptions is how a $100 film turns into a $300 one.
  • Start on free tiers. Prototype your look and characters before you pay for a month.

Not sure which tools fit your style and budget? The free Pipeline Picker recommends a stack in five questions and estimates the monthly total.

Don't cheap out on rights

If you plan to put your film in festivals or monetize it anywhere, check what each tool's plan allows. I used a paid Suno tier for the score because that's what covers commercial use. Free tiers on many tools are limited to personal use. Read the terms of the plan you're actually on. If you clone a voice, make sure it's your own, or get written consent from the person. Festivals increasingly ask about this.

Festival fees are a separate budget

Making the film is one budget and getting it seen is another. Entry fees range from free (many AI-specific contests) to well over $50 for big festivals. I keep a separate cap for entries. If you're heading that way, read my walkthrough on how to submit an AI film to festivals, or browse the free Festival Finder.

Want the full breakdown?

I packaged the whole Seahorses workflow into a guide: the cost and time sheet, the reject → lock loop, the prompt playbook, the voice process and the upscale-to-festival-cut steps. It's $18. The AI Film Festival Submission Kit is $18, or you can get both in the bundle for $28.

SEE THE GUIDE & KIT →WATCH THE FILMS

FAQ

How much does it cost to make an AI short film?

For a 2–4 minute short, expect roughly $50–$300 in new subscriptions and credits if you already own an editor. My 3-minute horror short All the Pretty Seahorses cost about $100. My tech-noir short FEMTECH cost about $200.

How long does it take to make an AI short film?

My two 3-minute shorts took about 24 hours over 4 days and 20–30 hours over 4–5 days. Most of that time goes to judging takes, fixing continuity and editing, not waiting on renders.

What is the biggest cost in AI filmmaking?

Video generation, especially rejected takes. You pay for every clip you throw away, so locking characters early and drafting at low resolution saves the most money.

Can I make an AI short film for free?

You can prototype a look, characters and a short teaser on free tiers. A finished multi-minute short with consistent characters, and the commercial rights to put it in festivals, usually needs at least one paid plan.

← All Learn articles