Compare AI Figure Credits by Cost per Finished Figure
A worked spreadsheet example separates subscription fees, generation attempts, revisions and manual cleanup so credit balances become useful budget numbers.
A plan with 1,000 credits is not necessarily better value than a plan with 200. Credits are internal units: one provider may charge per generation, another per image size, and another separately for editing. The useful denominator is the number of figures your team can actually finish.
This article uses hypothetical plans, not current prices for named products. Replace the inputs with the provider’s checkout price and credit rules before making a purchase. The arithmetic is deliberately visible so you can reuse it.
Turn the credit rules into a workload
Suppose your lab needs eight finished mechanism diagrams this month. In a small pilot, each accepted diagram takes three initial generation attempts and two local edits. Your editor also spends 15 minutes arranging labels and checking the export.
Two fictional plans have these rules:
| Input | Plan A | Plan B |
|---|---|---|
| Monthly subscription | $24 | $36 |
| Included credits | 120 | 240 |
| Initial generation | 4 credits | 6 credits |
| Local edit | 2 credits | 3 credits |
| Additional credit pack | 40 credits for $12 | 60 credits for $12 |
The per-figure credit demand is 3 Ă— generation + 2 Ă— edit:
- Plan A:
3 Ă— 4 + 2 Ă— 2 = 16credits. - Plan B:
3 Ă— 6 + 2 Ă— 3 = 24credits.
Eight figures therefore need 128 credits on A and 192 on B. A requires one additional pack, even though only eight credits exceed its allowance. B fits within its included credits.
| Monthly calculation | Plan A | Plan B |
|---|---|---|
| Credits required | 128 | 192 |
| Additional packs required | 1 | 0 |
| Cash paid | $36 | $36 |
| Cash cost per finished figure | $4.50 | $4.50 |
| Unused credits after the work | 32 | 48 |
The larger credit balance did not produce a lower cost here. Conversely, the cheaper headline subscription did not produce a cheaper month. Pack size and the actual number of attempts changed the comparison.
Reuse these spreadsheet formulas
Create one row per candidate plan. Give the columns meaningful names rather than relying on an informal conversion like “one credit is ten cents.”
CreditsPerFigure = DraftAttempts * CreditsPerDraft
+ EditAttempts * CreditsPerEdit
MonthlyDemand = FinishedFigures * CreditsPerFigure
ExtraPacks = ROUNDUP(MAX(0, MonthlyDemand - IncludedCredits)
/ CreditsPerPack, 0)
MonthlyCash = SubscriptionPrice + ExtraPacks * PackPrice
CashPerFigure = MonthlyCash / FinishedFigures
If a provider has no top-up packs, replace that calculation with the next eligible subscription tier. If different resolutions have different charges, give each image type its own workload row. Do not average a cheap thumbnail together with a high-resolution poster unless that reflects the work you really expect.
Run a sensitivity test before committing
Your first pilot may be unusually easy. Repeat the calculation with five drafts and four edits per figure. Under the same fictional rules, A needs 28 credits per finished figure and B needs 42. Across eight figures:
| Harder revision workload | Plan A | Plan B |
|---|---|---|
| Credits required | 224 | 336 |
| Additional packs | 3 | 2 |
| Monthly cash | $60 | $60 |
| Cash per finished figure | $7.50 | $7.50 |
This example happens to tie again because of the selected pack prices. Your real plans may cross at a particular workload. Calculate at two, eight and twenty finished figures rather than declaring one universal winner.
Also record why drafts failed. An image with the wrong arrow direction and an image with unreadable labels may need different repair strategies. If all extra attempts fix text, adding labels manually could be cheaper than buying more generations. If the scientific topology is repeatedly wrong, a clearer brief or a different drawing workflow may matter more than credit price.
Include the time spent after generation
At an illustrative internal labor rate of $30 per hour, 15 minutes of cleanup costs $7.50 per figure. In the first scenario, the combined figure cost is therefore $4.50 + $7.50 = $12, excluding scientific review time shared by both workflows.
If another workflow takes 35 minutes of cleanup, that portion becomes $17.50. A two-dollar saving on generation would not compensate for the extra ten dollars of labor. Track time to fix labels, redraw a relationship, align panels and produce the required export separately; otherwise “AI saved time” remains an impression.
What to record during a three-figure pilot
Use one easy diagram, one typical mechanism and one difficult multipanel figure. Keep the same brief for each candidate. Log the starting balance, each charged action, any refunded failure, elapsed manual editing time and whether the final file can be revised by a coauthor.
A useful log row is:
Figure: membrane transport mechanism
Action: revise the direction of the efflux arrow
Balance before: 88 | after: 86 | charged: 2
Outcome: arrow fixed; labels preserved
Manual cleanup after action: 4 minutes
Do not infer a refund from an error toast; check the balance or transaction record. Record expiry and rollover rules because unused credits may have no value next month.
The decision is then concrete: choose the plan that supports your expected workload, required output and revision time at an acceptable total cost. A large credit number alone answers none of those questions.