$50 of 6-second vertical ad tests: 66 on Wan, 43 on Mini, 14 on 2.5

Fifty dollars buys 66 six-second 9:16 clips at 720p on Wan 3.0, 43 on Seedance 2.0 Mini, or 14 on Seedance 2.5. Each count is floor(5,000 cents divided by the clip price), using the billable prices as of 2026-10-08. This is the number to pick a model by when the goal is to find a hook, not to make the final cut.

Counts

Wan 3.0 accepts 2 to 30 seconds, Seedance 2.0 Mini 4 to 15 seconds, and Seedance 2.5 4 to 30 seconds, so 6 s is legal everywhere. The leftover is what the last full clip does not use.

Arithmetic

Wan: 5,000 / 75 = 66.7, so 66 clips, 66 x 75 = 4,950 cents.

Mini: 5,000 / 114 = 43.9, so 43 clips, 43 x 114 = 4,902 cents.

Seedance 2.5: 5,000 / 347 = 14.4, so 14 clips, 14 x 347 = 4,858 cents.

Dropping to 480p

Every row also lists 480p. A 6 s Wan 3.0 clip at 480p is $0.38, so $50 buys 131 of them (131 x 38 = 4,978 cents). Seedance 2.0 Mini at 480p 9:16 is $0.53 for 6 s, which gives 94 clips (94 x 53 = 4,982 cents).

A TikTok non-Spark ad needs at least 540x960 pixels for 9:16. Sume's docs name resolutions by label rather than by pixel size, so probe one 480p result with video inspect before you plan on it for an ad. The Spark Ads overview page lists no resolution restriction, which makes a draft tier less risky for a Spark test than for a standard ad.

Using the cheap tier

Draft cheap, then spend on the winner. If 66 drafts on Wan surface three hooks that hold attention, rerender those three on the model you want to ship, at the same 6 seconds. Three reruns on Seedance 2.5 cost 3 x $3.47 = $10.41, so the total is $49.50 + $10.41 = $59.91.

Sume does not guarantee that the same prompt gives a similar clip on a different model, so the rerun is a new creative decision, not a pixel-exact upgrade.

How the counts are reached

A budget count is the budget divided by the clip price, rounded down, because a partial clip cannot be bought. Wan 3.0 at 720p for 6 seconds is 75 cents, so 5000 / 75 = 66.7, which is 66 clips. The Seedance 2.0 Mini 9:16 row is 114 cents, so 5000 / 114 = 43.9, which is 43 clips. Seedance 2.5 is 347 cents, so 5000 / 347 = 14.4, which is 14 clips.

The leftover matters if you run in rounds. Wan leaves 5000 - 66 x 75 = 50 cents, Mini leaves 5000 - 43 x 114 = 98 cents and Seedance 2.5 leaves 5000 - 14 x 347 = 142 cents. None of these is enough for another clip of the same kind, so the leftover is not worth chasing.

Using the count

Spark Ads need an existing organic post, and TikTok's Spark overview limits an account to 10,000 Spark Ads, so test volume is not the constraint here; spend is. A 66-clip batch on Wan is one clip each for 66 hooks, which is more variety per dollar but one visual style. The 14 clips on Seedance 2.5 are fewer but the model takes up to 30 seconds, so the same budget buys shorter tests only.

Decide the question first. If the test is about the hook words, a cheap model is enough and the clip count matters. If the test is about visual quality, spend on fewer clips. Neither TikTok page ranks models, and these counts do not either; they are only arithmetic on the catalog prices.