What a Finished Minute of AI Video Really Costs

Most AI video tools price by credits, by generation or by second of clip. Those numbers are real, but they answer the wrong question. A creator does not want a clip. They want a finished video. The honest question is: what does a finished minute cost, once everything it took to get there is counted?

This guide breaks that down.

Why a finished minute is not one clip

A minute of narrated long-form video is not one sixty-second generation. It is a sequence of scenes, each illustrating a beat of the narration. In a typical narrated video, a scene lasts somewhere between a few seconds and a quarter of a minute, so one finished minute can contain many separate shots.

For each of those shots, the work behind it may include:

• writing and checking a shot description
• drawing several still candidates
• judging those stills
• writing a motion prompt
• generating one or more video takes
• checking the clip

Only one clip per scene ends up on screen. Everything else is still part of the cost.

The cost drivers

1. Scenes per minute

More scenes per minute means more shots to describe, draw, animate and check. Fast-cut styles cost more than slower, narration-led pacing. Letting scene length follow the narration, rather than cutting on a fixed rhythm, keeps the count sensible.

2. Retakes

Retakes are usually the largest hidden cost. A shot that takes four attempts costs roughly four times as much as one that works first time, and the three failures never appear in the video. Retakes come from:

• characters or places drifting from their references
• the picture not matching the narration
• hands, faces or objects distorting in motion
• ambitious camera moves or crowded action

Anything that raises the first-time success rate lowers the cost of a finished minute more than any discount on generation.

3. Stills versus motion

Video generation generally costs much more than still-image generation. That makes stills a cheap place to catch problems. If you approve the picture first and animate from it, you pay for motion only on shots whose composition, characters and location are already correct. Skipping stills and generating video directly from text tends to push the failures into the expensive step.

4. Judging and checks

Checking a shot description against the narration, and checking a still for identity, location, props and framing, costs something. It is small compared to a failed video take, and it is what prevents those failed takes. Cutting checks to save money usually costs more in retakes.

5. Narration

Narration is part of the finished minute. Voice generation, timing it against scenes, and re-recording lines that change all belong in the total.

6. Setup, spread across the video

Building character sheets, location references and a style guide is a one-time cost per production. Across a ten-minute video it is spread thin. Across a series that reuses the same world, it gets cheaper per minute with every episode.

7. Your own time

The cost that never shows on an invoice is the hours spent prompting, sorting takes, fixing continuity and moving files between tools. For many creators this is the real limit on how often they can publish.

Why “per finished minute” is the honest unit

Credits, tokens and per-clip prices put the risk of failure on the creator. If a shot takes six tries, you pay six times, and the price list never told you to expect that.

Pricing per finished minute puts the unit where the value is: video you actually keep. It also makes costs comparable across tools, because it forces every hidden step (stills, checks, retakes, narration) into the same number.

How to estimate your own cost per finished minute

1. Count the scenes in a typical minute of your video.
2. Estimate the stills you draw per scene before choosing one.
3. Estimate the video takes per scene before one is kept.
4. Add the cost of checks and narration.
5. Add your own hours at a rate you would pay someone else.
6. Divide the total by finished minutes, not by clips generated.

Run that on your last project. The result is usually well above what the per-clip price suggested, because the failures and checks were never on the price list.

A rough example of the arithmetic, with no prices attached: if a minute has eight scenes, you draw three stills per scene and keep one, and each scene needs two video takes on average, that minute contains twenty-four stills, sixteen clips, eight shot checks and a minute of narration. Only eight clips reach the screen.

How to bring the cost down

• Lock the script before generating anything.
• Build a production bible so shots start from fixed references.
• Check every shot description against the narration before drawing.
• Approve stills before paying for motion.
• Keep motion modest; complex multi-person action and hands are where retakes pile up.
• Use one consistent model choice per type of shot rather than experimenting on every scene.
• Review the whole film once, then fix only what visibly fails.

How VIA prices it

VIA is built around this unit. Plans include a number of finished video minutes per month, with extra minutes as top-ups. You do not buy credits per generation or pick models per shot; VIA chooses the image and video model for each shot and runs the checks that keep retakes down. VIA is in private beta with beta pricing. See pricing for the current plans, and how it works for what happens inside each finished minute.

If you want to know what a finished minute costs before you start, rather than after, apply for the beta.