Six-shot 30-second ad budget by shot role on Sume

A 30-second ad built from six five-second shots costs $4.45 on Sume if only the hero shot uses Seedance 2.5 at 720p, the two product shots use MiniMax H3 Max at 768p and the three cutaways use Grok Imagine video with a low-quality GPT Image 2.5 still as the start frame. Making all six shots Seedance 2.5 at 720p costs $17.63. The mixed plan is 25% of the all-hero plan.

The saving is real money, but whether b-roll at about $0.087 looks right next to a $2.889 hero is a creative judgment this post cannot make for you. Test the pairing on one ad before you commit a campaign to it.

Which shots need the best model?

Think in roles. The hero shot carries the idea and gets seen the longest. Product shots must show the thing you sell accurately. Cutaways are texture between the other two: a street, a hand, a window. Spend by role instead of by default, and send each shot to a model whose catalog entry supports the duration, resolution and references you need. Remember the docs list different ranges: every catalog model other than a few long-form ones tops out at 15 seconds, while Seedance 2.5 accepts 4 to 30.

What does the mixed plan cost line by line?

Prices are the customer estimates from Sume's pricing package for five-second requests: Seedance 2.5 at 720p 9:16, H3 Max at 768p, Grok at 720p. Grok Imagine video on Sume requires one input image, so each cutaway also pays for a low-quality GPT Image 2.5 still at $0.02475. Timeline render is $0.10 per output minute, rounded up per the timeline docs, so a 30-second ad bills one minute. Captions are a flat $0.20 per job on videos up to 60 seconds.

Where does the money go?

In the mixed plan the hero shot is $2.889 of $4.451, or 65%. Even after you go all-in on cheap b-roll, one premium shot dominates the bill, so that is where a retry budget belongs. If the hero takes three tries, the ad costs $4.451 plus $5.778, or $10.23, still below the all-hero plan before any of its retries.

Captions and the render add $0.30, 7% of the mixed plan. They are fixed costs per ad, so they matter more as the clips get cheaper.

How to apply this to a batch of ads

Write the shot list with a role next to every shot before you generate anything. Then set the model and the resolution per role in your request builder, not once for the whole job.

• Keep one function per role that returns the exact request body, so the price of a shot is predictable.
• Generate the hero first; if it fails, nothing else is worth rendering.
• Run the cutaways at the cheapest resolution that survives a phone screen.
• Compare against the all-hero total before each campaign and write the delta down; if the cheap plan converts worse, the saving did not matter.
• Use the unbilled POST /v1/timeline-1.0/plan to check the render before you pay for it.