The arithmetic nobody puts in the demo reel

The number that decides your project

Suppose a shot has to hold a character’s face convincingly, and your generations do that acceptably seven times out of ten. That feels fine.

Now you need eight consecutive shots that all match. The probability that all eight land, if attempts are roughly independent, is 0.7 to the power of 8, which is about 0.06. Six per cent.

You will not get eight in a row. You will generate each shot until it works, which is the real workflow, but the arithmetic tells you the cost: at a 0.7 hit rate you generate on average 1.4 takes per shot, so eleven or twelve generations for the sequence, plus the ones you reject for other reasons.

Drop the hit rate to 0.4, which is realistic once a shot has a specific requirement, and it is two and a half takes per shot, twenty generations, and a meaningfully worse chance that any of them are actually good rather than merely acceptable.

Measure your own rate

The number above is invented. Yours is not, and it takes a week of normal work to find.

Keep a tally. For every shot: how many generations before one you would ship, and what the rejections were for.

s04  3 takes  (1 face drift, 1 hand)
s05  1 take
s06  6 takes  (3 face drift, 2 background churn, 1 too much motion)
s07  2 takes  (1 wardrobe colour)

After twenty shots you have two things: a hit rate you can quote in a budget, and a ranked list of what actually causes your rejections. Most people are surprised by the second. The thing you worry about is rarely the thing that costs you.

Reduce the exposure, not the rate

You have limited control over the per-shot rate. You have complete control over how many shots have to match.

This is the professional response and it is why it keeps appearing in this course:

• Fewer shots of the character. A three-shot scene where the face is clearly visible once is far more achievable than a six-shot scene where it is visible every time.
• Over-the-shoulder and back-of-head shots. The face is not in them, so it cannot drift.
• Wides where the face is forty pixels across. Drift is present and invisible.
• Cutaways. Hands, objects, the thing being looked at, a detail of the room. Each one is a shot that carries the scene without exposing the hard thing.
• Reaction cut to something other than the character.

Every one of those is also, independently, better editing. A scene told entirely in shots of a face is weak filmmaking regardless of how it was produced.

Where the rate improves

In order of effect:

1. A trained identity model. The largest single improvement available.
2. First frames derived from a character sheet rather than described.
3. Shorter shots. Drift grows with distance from the anchor, so a four-second shot holds where an eight-second one does not.
4. Lower motion. Fewer frames of change, fewer chances to break.
5. Reference conditioning. Real, and weaker than the above.

Notice that three of the five are decisions about the shot rather than about the model. That ratio is the lesson.

What this means for a quote

If you are billing, the hit rate is the number your price is built on, and it belongs in the conversation rather than hidden inside an hourly rate.

Quote in shots with a stated number of included attempts. “Thirty shots, up to four attempts each, further revisions billed per shot” is a sentence that protects both sides: it tells the client that attempts cost money, and it tells you what you agreed to when they ask for the twelfth version of shot nine.

The money module deals with this properly. It is raised here because the hit rate is the input.

The honest note about demos

Anyone showing a character-consistent, ten-shot narrative without mentioning a discard pile is showing you the survivors. That is not dishonesty exactly (it is a showreel, and showreels have always been the top one per cent), but it sets an expectation that costs beginners real money and real confidence.

The rate is not a measure of your skill. It is a property of the tool, and the skill is in designing work that does not depend on it being high.

Today: start the tally. Twenty shots from now you will have a number you can put in a quote, and a ranked list of what actually goes wrong in your work.