A $500 holiday video budget: how many seconds each Sume model buys

Five hundred dollars buys 8,000 seconds of Wan 3.0 at 480p, 4,000 seconds at 720p and 2,000 seconds at 1080p. On Avatar 1.0 plus it buys about 2,040 seconds of a talking presenter, and on Gemini Omni Flash 1.1 at 360p it buys about 13,333 seconds.

A seconds budget is the useful way to plan holiday video, since the ad count follows from it. Divide by the length of an ad and you have the number of variants.

Seconds per $500

Sume's catalog models are billed at the provider list times 1.25 per output second. Avatar 1.0 uses its charged per-second rates by tier. The table divides $500 by the rate.

How to split it

Do not spend all of it on finals. A good split for a hook-test-led holiday plan is 20 percent on drafts, 50 percent on finals of the winners and 30 percent held back for late changes: a price cut, a stock-out, a new offer. On $500 that is $100, $250 and $150.

The held-back share is the one people skip. Black Friday offers change in the last week, and a plan with no reserve turns every change into an approval meeting.

These figures are render costs only. Image edits, voice, music, assembly and caption jobs are small next to video but not zero: Timeline assembly is $0.10 per started output minute, and a video trim is $0.02 flat. Add them to the plan.

From dollars to ads

Convert the plan into counts. With $100 of drafts at Wan 3.0 480p, you get 1,600 seconds, which is 160 ten-second drafts. With $250 on finals at 720p Wan, you get 2,000 seconds, which is 200 ten-second finals or 66 half-minute ones. These are far more than most teams can review, so the constraint is people, not money.

The honest message: model cost is rarely the limit on a holiday video plan. Review time, approvals and testing windows are. Use the seconds table to show that to a finance team that expects a video budget to be large.

Check before you commit

The numbers move. Read GET /v1/videos/models for current prices before you commit. Provider list prices can change, and Sume bills the current list times 1.25 at the time of the job.

A worked plan for three products

Here is a worked plan for a small store with three products and one offer. Each product gets four hook variants, which is twelve drafts of 10 seconds on Wan 3.0 at 480p. That is 120 seconds and costs $7.50. The team picks the top three hooks and renders each at 720p for 15 seconds, which is 45 seconds and $5.63 at $0.125 per second. Then the best of those is rendered at 1080p at 15 seconds for $3.75.

The total is $16.88 for a full test-and-final cycle on one offer. Even if you repeat it for five offers over the season, you are at about $85. A $500 budget would still have $415 left over for avatar clips, image edits and assembly.

If the same plan used Avatar 1.0 plus presenter clips, 12 clips of 10 seconds is 120 seconds at $0.245 per second, or $29.40. Presenter clips are dearer per second but replace a shoot, a booking and an edit.

Rules that keep the budget real

Two rules keep the plan honest. First, count output seconds, not clip counts, since the bill follows seconds. A failed or rejected job does not change that: if you re-render, you pay for the new seconds. Second, price the cheapest draft tier into every test, so that a no-go idea costs cents rather than dollars.

Finally, set a ceiling. Spend caps on Format runs and queues let you fix the most a batch can cost before it starts, so a mistaken bulk submit cannot spend the reserve. Write the ceiling down next to the budget split.