What a Minute of AI Video Actually Costs
Per-clip list price is the least important number in your bill. A minute of finished video is 10–20 shots, and cost is driven by how many times you generate each shot. Three attempts per shot means paying for three minutes to deliver one. Cut the multiplier instead of hunting a cheaper model: approve the script and the stills before you render video, and route only hero shots to premium models.
The multiplier, not the rate
Everyone comparing AI video tools compares the cost per generation. It’s the wrong column. Two tools at the same rate produce wildly different bills depending on how many rejected takes it takes to get a usable shot, and a rejected take is billed exactly like an accepted one.
Do the arithmetic on a modest project. A three-minute short is somewhere around 40 shots. At 1.5 attempts per shot you pay for 60 generations. At 3.5 — entirely normal when you render before the script is settled — you pay for 140. Same short, same models, more than double the cost, and the difference is entirely process.
Where the waste actually happens
Overspend concentrates in a specific place: generating video to evaluate a decision that isn’t about video. Rendering a scene to find out whether the dialogue works. Rendering a close-up to find out whether you like the character’s face. Rendering the whole sequence to discover the scene runs long.
Each of those questions has a much cheaper instrument, and every one you move downstream gets answered at a fraction of the price.
The cost ladder
Order your production so each decision is made at the cheapest stage that can answer it. Roughly, ascending:
- Text — effectively free. Script, structure, dialogue, shot list. Every story problem should die here. A scene that doesn’t work on the page will not be rescued by rendering.
- Still images — cents. Character identity, wardrobe, location plates, and a storyboard frame for every shot. This is where look, framing, and continuity get approved.
- Voice — cheap. Record the dialogue and listen to it. Now you know the real runtime of every scene before a single video credit is spent.
- Still-motion cut — cheap. Assemble the stills with camera moves over the finished voice track. You can watch the entire piece and fix pacing, order, and length for almost nothing.
- Video generation — the expensive part. Only now, and only for shots whose framing, cast, location, and duration are already approved.
The highest-leverage habit. Watch the whole thing as a still-motion cut before you render any video. Almost every expensive mistake — a scene that drags, a shot that isn’t needed, an act that’s in the wrong order — is visible there and costs nothing to fix.
Route models per shot, not per project
Video models differ in price by a large multiple, and the quality gap between them is wildly uneven across shot types. On a close-up with dialogue and subtle expression, the premium model is obviously better. On an establishing wide, a cutaway to a phone screen, or a two-second insert of a hand on a door handle, the cheap model is indistinguishable in a finished cut.
Picking one model for a whole project therefore either overpays on most shots or underdelivers on the few that matter. Per-shot routing — premium for heroes, economy for everything else — is one of the largest single savings available, and it costs nothing in quality because it’s spending the money where it shows.
Read the expiry terms
Most credit systems expire monthly allocations without rollover. That makes an idle month a total loss and a heavy month a top-up purchase, so the effective rate is worse than the advertised one unless your output is steady. Before committing, check three things: whether credits roll over, whether unused credits expire, and whether top-ups require an active plan.
How ShowMaker Studio handles it
The pipeline is built as that ladder. The Writers’ Room settles the script before anything is generated; character and location stills are approved and locked before video; the Listening Room settles the voice performance; and a narrated still-motion cut lets you watch the whole show before buying a single video render. Shots route per-shot to the model that suits them rather than per-project, and paid renders auto-lock so an approved take isn’t accidentally regenerated. Current plans and credit costs are on the pricing page, and the free tier covers narrated still-motion films so you can complete a piece end-to-end before spending anything.
Related questions
How much does a minute of AI video cost?
The honest answer is that the model's list price is the smaller half of the bill. A minute of finished video is roughly 10 to 20 shots, and what determines your cost is how many times you generate each of those shots. At three attempts per shot — normal without approval gates — you pay for three minutes of generation to deliver one. Cutting re-renders is worth more than finding a cheaper model.
Why do my AI video credits disappear so fast?
Because rejected renders cost the same as accepted ones. Every discarded take is fully billed, so a project where half the shots are re-rolled twice costs double for identical output. The spend concentrates in shots you generate before you know whether the script, the character look, or the framing is right.
What's the cheapest way to make AI video?
Approve everything cheap before you buy anything expensive. Lock the script, then the character and location stills, then generate video. Still images cost a small fraction of video, so every decision moved from the video stage to the still stage is a large saving. A narrated still-motion cut also lets you watch the whole piece for pacing before spending on generation at all.
Should I use the most expensive video model?
Only where it shows. Premium models earn their price on close-ups with dialogue, complex motion, and hero shots. Establishing wides, cutaways, inserts, and B-roll look effectively identical from a cheap model. Routing per shot rather than per project typically cuts a project's render cost several times over with no visible quality difference.
Is a subscription or pay-per-render better for AI video?
It depends on whether your work is continuous. Subscription buckets are cheaper per credit and suit ongoing production, but monthly credits usually expire, so an unused month is money lost. Pay-per-use suits occasional projects. The thing to check on any plan is expiry and rollover — that's where the real cost difference hides.
How do I estimate a project's cost before starting?
Count shots, not minutes. Break the script into scenes, estimate shots per scene (a dialogue scene is typically 6 to 12), multiply by your tool's per-shot render cost, and then multiply by your realistic attempts-per-shot — 1.5 with approval gates, 3 or more without. That last multiplier is the number you control.
Every fix on this page is a feature, not a workaround.
ShowMaker Studio is built around these problems — reference-locked casting, location look signatures, coverage-based scenes, delivery-timed captions, and per-shot model routing. Three free films a month, no card.