0.4 The Break-Even Calculator: Before You Spend a Dollar

Most people in this business cannot say what one published post costs them. They know their subscription total, roughly, and they know it feels like a lot in a bad month. That is not a number, and without a number every tool decision is a guess.

Here is the number. It takes four minutes and it will change what you buy.

The formula

cost per published post = (monthly money cost + monthly time cost) / posts published per month

Two decisions follow from the result, and they are the reason the formula is worth doing:

• Below $6 per post — healthy. Keep going, and consider stepping up a tier.
• Above $15 per post — step down. You are paying for capacity you are not using.
• Between them — it depends on what you are buying with the difference. See the retries section below, because this band is where the subscription-versus-metered decision is made.

The $6 and $15 thresholds come from practitioner cost accounting across the threads in this research, not from a vendor. [UNVERIFIED] — the sample is small and self-reported. They are useful as a sanity check, not as a law.

Step one: the money

Three real stack tiers, as observed in October 2026:

Practitioner-reported cost per published post lands between $1.80 and $10.60 across these tiers. That range is wide because the tiers are not the variable that matters — the number of posts is.

The trap inside the tiers is how the vendor charges you:

• A subscription absorbs retries. Generating thirty candidates and publishing one costs the same as generating one, in money. The cost is your time.
• A metered API multiplies retries. Thirty candidates and one published post is thirty times the money.

This single distinction decides more of your budget than your choice of model does, and it is where the chapter on pricing (see 5.6) earns its place.

Step two: the hidden multiplier

Identity drift means you will not publish your first generation. Practitioners report needing 2 to 5× the generations they expected in order to get one usable output, because the persona drifts on pose changes, in closeups, and in full-body frames.

Apply the multiplier to whichever cost model applies to you:

If your stack is metered, a $5.00-per-post estimate is a $15.00-per-post reality. That is the difference between the healthy band and the step-down band, and it is invisible until you count generations rather than dollars.

Step three: the time, priced honestly

Three to four hours a week is the realistic production load for a single persona, which is 13 to 17 hours a month. Price that at whatever an hour of your work is worth. At $25 an hour:

money:      $102
time:       17 h × $25 = $425
total:      $527
posts:      20
cost/post:  $26.35

The time cost dominates the money cost by a factor of four. This is the most important sentence in Part 0. Everyone tunes the $102. Almost nobody prices the $425, and it is the number that decides whether this is a business or an expensive hobby.

The consequence is not that you should not do it. It is that the only way to make the arithmetic work is to lower the hours per post, not the dollars per tool. A workflow that cuts production from 17 hours to 6 hours a month is worth more than any subscription downgrade, and it is what Parts IV, V, and VII exist to build.

If you price your time at zero because it is your own time, be explicit about that choice rather than hiding it inside the formula. There is no wrong answer here. There is only an unexamined one.

Hardware, if you buy it

A local GPU changes the shape of the calculation. A used RTX 4090 with 24 GB runs roughly $0.50 to $2.00 per hour of equivalent cloud GPU time, and practitioners report a 6-second image-to-video render taking one to two minutes locally.

The case for buying is volume. The case against is that you are pricing a capital expense against a rental, and the rental keeps getting cheaper. Chapter 3.3 works through the threshold with the VRAM requirements, and its conclusion is blunt: 24 GB is the realistic floor and 16 GB is not enough, so the cheap card you are considering will not run the workflow this book teaches.

Where this goes wrong

Counting only the subscriptions. The four-figure line is time. Ignoring it produces people who believe they run a profitable operation at $1.20 per post while working a second unpaid job.

Forgetting the multiplier. A metered stack without a retry allowance in the budget is a stack that surprises you in month two, which is when most abandonments happen.

Downgrading the wrong thing. Cutting the image subscription to save $22 while adding three hours a month of manual work is a $22 saving that costs $75. Compare hours first, dollars second.

Treating the thresholds as targets. $5.99 per post is not a goal. It is a ceiling that tells you when to stop spending, not a level to reach.

The gate