Appendix D — Calculators and Worksheets

Four calculations decide most of the money questions in this book: what a post costs, how many subscribers a persona needs, what a minute of video costs, and when self-hosting replaces a metered API. Each one is written for a spreadsheet rather than for a script, and each comes with one worked example on the chapter's own numbers.

The formulas are deliberately shallow. Every term is a line you can find on a statement or a dashboard, and the answer is only as good as the input you least enjoy collecting. Where an input is an estimate, write "estimate" beside it, because an unlabeled guess is how a spreadsheet starts lying.

Convention: Every input is point-in-time and carries the chapter's [AS OF 2026-10] date. Outputs are estimates of a decision rather than forecasts of a month, and the chapter cited with each calculator holds the reasoning the formula compresses.

1. Cost per published post

The number that decides your tier, and the one that retries distort most. Generate a hundred frames to publish four posts and every rejection is a line in the numerator while the denominator stays at four.

cost_per_post = (pipeline + promotion + tooling) / posts_published

pipeline        generation, voice, video, and lip-sync for the period
promotion       paid promotion attached to those posts
tooling         subscriptions not already counted in pipeline
posts_published assets that actually went live

Worked example, one month, four posts published from 150 frames at a one-in-three survival rate. On a subscription, the tier's monthly allocation is about $36, so $36 ÷ 4 = $9 per published post, plus your time. On metered per-image billing, the same 150 frames at roughly $0.36 each cost $54, so $54 ÷ 4 = $13.50 per published post.

What it means: the gap between the two billing models is the cost of failure, not the cost of quality. Chapter 0.4's thresholds read below roughly $6 as healthy and above roughly $15 as a signal to step down a tier, with the middle band running from about $1.80 to $10.60. If your figure sits above the band, change the survival rate or the billing model before you change the model.

2. Persona break-even

How many paying subscribers one persona needs before it stops costing you money. Two answers matter: the subscription-only answer, and the blended answer once chat revenue is included.

subscribers_needed = fixed_monthly / margin_per_subscriber

fixed_monthly  pipeline tier + chat tooling + storage
margin_per_sub price - payment and platform fees - delivery cost

Worked example on the Working tier. Fixed cost is the tier's roughly $102 plus about $60 of tooling, so $162. At a $10 entry price with ten percent in fees, the contribution margin is $9, giving 18 subscribers. Substituting the blended ARPU from chapter 10.4 — $70.50, because PPV and tips carry most of the revenue — gives 2.3 subscribers, which rounds to three.

What it means: the subscription-only figure is the honest one for a persona with no chat layer, and the blended figure is the one that applies once chat exists. Run both against each tier: Minimum at roughly $36 needs four subscribers on margin, Working needs 18, Volume at roughly $212 needs 31. A persona that cannot reach the subscription-only figure within two quarters has a positioning problem rather than a budget problem.

3. Cost per minute of video

Video is where the marginal cost lives, and the rejection factor is the term operators leave out.

cash_per_minute = (seconds * price_per_second * rejection_factor)
                  + voice + lipsync

cash_per_minute  divided by finished minutes, not generated minutes
voice            per minute of speech delivered
lipsync          per minute of synced footage
rejection_factor unusable clips per usable clip, as a multiplier

Worked example, one 45-second video at $0.20 per second with a rejection factor of two. Generation is 45 × $0.20 × 2 = $18.00. Voice for 45 seconds is about $0.20; lip-sync at roughly $4 per minute is 0.75 × $4 = $3.00. Cash total $21.20, or about $28 per finished minute. Adding 45 minutes of editing at $25 per hour gives $18.75 more, for $39.95 all in.

What it means: the cash figure is what the card is charged and the all-in figure is what the persona costs. Chapter 6.4's band of roughly $25 to $30 per finished minute on the standard tier holds at this rate, and the Lite tier changes the picture sharply — the same clip generated at $0.05 per second lands near $10 per finished minute in cash, which is why volume plans run there and hero shots do not.

4. Self-host versus API

The decision is a crossover, not a preference. Fixed cost on one side, per-image cost on the other.

monthly_api  = images * price_per_image
monthly_self = gpu_hourly * hours + storage
crossover    = monthly_self / price_per_image

Worked example on three volumes, with a rented 4090-class machine at $1.00 per hour for 40 hours plus $10 of storage, so $50 per month fixed.

The crossover against metered billing is $50 ÷ $0.36 = 139 images per month. Below that, the API is cheaper; above it, the rented GPU is. Against a flat subscription the crossover is $50 ÷ $0.006 = 8,333 images, which is why the subscription is the cheapest line at every realistic solo volume and why the real reason to self-host is not price.

What it means: at 2,000 images a month, self-hosting beats metered billing by more than $600 and loses to a $10 subscription by $38 — but the subscription absorbs retries and the GPU absorbs nothing you do not configure. Choose between them on continuity and portability, per chapter 12.4, and let the arithmetic confirm that you are not paying a premium for control at a volume where control buys nothing.

Worksheet: the inputs to collect each month

Fourteen numbers feed all four calculators, and thirteen of them are already produced by the pipeline or the platform dashboard. Collect them once a month in the same order, and date the page.

What it means: two of these are the ones operators avoid, and they are the two that change decisions. Production hours turn a $9 post into a $30 post, so a persona that looks profitable per post can be the worst use of your week. Cancellations separately from subscribers turn a growing headline into a churn number, and churn is what decides whether the persona compounds or merely repeats.

Reading these numbers honestly

Three cautions. Every input here is point-in-time, and the ones that move fastest are per-second video prices and voice quotas, so re-run these after any vendor change rather than adjusting a figure in your head. Every output is an estimate of which decision is better, not a forecast of what next month will cost.

And each calculator answers only the question it was given. Cost per published post says nothing about whether the post earned anything; break-even says nothing about whether subscribers will stay; cost per minute says nothing about whether the video was worth making. Use them to compare two options at the same volume, which is what they are built for, and check the result against the quarterly audit before acting on it.