How to read the gap between Seedance Fast and Mini

Updated 2026-10-02

Seedance 2.0 Fast and 2.0 Mini are both positioned as the lower-cost end of the family, so the gap between them is small compared with the gap to the full model. A small gap is easy to misread: a headline percentage can flip sign depending on which hosts and tiers you compare, and a per-second difference that looks trivial can be decisive over thousands of drafts. This page shows how to read the live numbers and then convert them into cost per accepted clip.

Read the table at the right granularity

The table lists each Seedance row with the cheapest and the priciest host and the tier it reports. Three traps are worth naming:

Per second, before VideoRouter's 2% platform fee. For tiered models each row compares the resolution tier with the widest host-to-host gap. Built 2026-10-02 from the live catalog.

1. Different tiers. Seedance is priced per second per resolution. If the Fast row reports one resolution and the Mini row another, the gap you see is partly a resolution gap. Match the tier first, using the model pages for rows that offer several.
2. Different host columns. Compare the cheapest host of Fast with the cheapest host of Mini, and the priciest with the priciest. Mixing columns makes the gap look larger or even reverses it. See why Seedance prices vary between hosts.
3. Rates move. A gap read last month may not hold today. Treat the table as current and the numbers in any older article, including your own notes, as stale.

Then compute the gap two ways. The absolute gap is rate(Fast) - rate(Mini) per second at the same tier and host column. The relative gap is that difference divided by rate(Mini). Use the absolute gap for budgets, because it multiplies by seconds; use the relative gap for sanity checks.

What the gap is worth at volume

Let S be the total requested seconds in a month. The monthly difference between running everything on Fast versus Mini is:

monthly_delta = (rate(Fast) - rate(Mini)) x S x f

where f is one plus the platform fee. For small S the delta is noise, and the choice should follow quality and workflow fit. For large S, such as mass drafting, it can exceed what you would save by optimising almost anything else, so it deserves a proper test.

From per second to per accepted clip

Completed jobs are billed whether or not you keep the result, while upstream failures are not billed. If k is the fraction of clips you accept for a variant on your own prompts, then:

cost_per_accepted(variant) = rate(variant) x L x f / k(variant)

Mini beats Fast per accepted clip when k(Mini) / k(Fast) > rate(Mini) / rate(Fast). Because the gap between the two is small, that ratio is close to one, and a modest difference in acceptance can flip the answer. This is the reason not to pick on rate alone: with a small rate gap, acceptance is the dominant term.

Draft maths: a cheap tier only helps if drafts are mostly throwaway

A draft workflow generates d exploratory clips on the cheaper variant per approved shot, then renders the shot once on a higher tier. The cost per approved shot is:

d x rate(draft) x L x f  +  rate(final) x L x f / k(final)

Changing the draft variant from Fast to Mini moves only the first term. It saves d x (rate(Fast) - rate(Mini)) x L x f, provided drafts on Mini give you the information you need in the same number of attempts d. If Mini needs more drafts to reach the same decision, the saving shrinks and can reverse. Count drafts per approved shot on both variants during a trial, not just clips accepted.

A short test that settles it

1. Choose 20 to 30 representative prompts and inputs from your real workload.
2. Run each on both variants at the same tier and duration, on a capped key.
3. Record acceptance and, for drafting, the attempts needed to reach a decision.
4. Plug the rates from the live table and your measured k and d into the formulas above, using the calculator for the unit costs.

With small samples, differences in acceptance are often within noise; in that case there is no evidence to prefer the higher-priced variant, and the lower rate is a reasonable default until you see otherwise. Whether either variant meets your quality bar is a separate judgement the price table cannot make.

Reporting the gap honestly

When you write the result up, state the tier, the host column, the date you read the table and the sample size behind any acceptance figure. "Mini is cheaper than Fast" is a claim that expires; "at this tier and host column on this date, the per-second difference was X and acceptance in our 25-prompt trial was similar" can be checked and updated. Keep the raw numbers next to the conclusion.

Where this fits in a plan

For the full staged workflow, see Mini vs Fast vs Full: a cost strategy for teams; for turning rates into per-minute figures, Seedance API cost per minute. The cost page has clip costs by length. To run the comparison, create a key and put a spend cap on it before you start.

Frequently asked questions

Keep reading

• How to Reduce Your Seedance API Cost — Five Practical Levers
• Seedance API Cost Per Minute: Budget Formula & Retry Math
• Seedance vs Kling API Pricing: A Fair Cost Comparison
• Seedance 2.0 Mini vs Fast vs Full: A Cost Strategy for Teams