What a faceless channel actually costs to run

The cost of a faceless video is not one number, it is four, and they behave very differently as the video gets longer. Knowing which is which tells you what to change when the bill looks wrong.

A 60 second reel, line by line

Roughly, for a one minute narrated video with nine scenes:

• Voiceover: about half the total, billed per character of script.
• Images: one per scene, a fraction of a cent each.
• Script: a single language model call, cents.

Voiceover dominates, and it is the only line that scales linearly with length. That one fact explains most of what follows.

Why cost per minute falls as videos get longer

Scene count does not grow forever. A sensible generator floors at around six scenes and caps at around fifteen, because a ten minute video with ninety images is a slideshow nobody watches. So images and script plateau while narration keeps scaling.

A ten minute video therefore costs roughly ten times a thirty second one in narration, but only about twice as much in images. Cost per minute drops as length rises.

Practical consequence: a flat price per video overcharges short videos and undercharges long ones. If a tool bills the same regardless of length, your short videos are subsidising somebody else.

Clipping is priced completely differently

Cutting shorts out of a long video has almost no per-clip cost. The expensive part is reading the source: the transcript is the prompt, and it is the same transcript whether you take one clip or ten.

So clipping cost follows the length of the source, not the number of clips. A longer source costs more to read, but it also yields more clips, so the cost per published clip keeps falling as the source gets longer.

This is why clipping is dramatically cheaper per published video than generating. If volume is the goal and you have long videos to work from, clip them.

Provider cost is not the only billing model

Everything above is the raw provider cost, which is what you pay on a bring-your-own-keys plan. Most tools instead sell you credits, and the two behave differently enough to change which is cheaper for you.

• On credits, a clip is usually a flat unit regardless of how long the source was. Predictable, and it stops a three hour podcast costing more than a twenty minute one.
• On your own keys, you pay the real provider cost, so a short source is cheap and a long one is not. Better if your sources are short or you already have volume discounts.

The rough rule: if you clip long sources at volume, flat credits usually win. If you generate a lot of short narrated videos, your own keys usually win, because narration is billed per character either way and you skip the margin.

Where people actually overspend

• Premium voice models on long videos. Narration is the biggest line, so voice tier matters more than anything else you can change.
• Regenerating. A video regenerated three times costs three times as much and is rarely three times better.
• Ten minute videos published daily, before knowing whether anyone watches past minute two.

A sensible starting budget

Start with 30 to 60 second videos, one a day, on a mid-tier voice. That is cheap enough to run for a month without thinking about it, and a month is roughly what you need before the numbers mean anything.

Scale length only after retention says people are staying. Scale volume only after one format is clearly working. Both are cheaper mistakes to make late than early.

Common questions