Omni Flash resolution ladder on Sume: 0.3x, 1x, 1.5x, 3x of 720p

On Sume, gemini-omni-flash-1.1 costs 0.3 times its 720p rate at 360p, 1.5 times at 1080p and 3 times at 4K. The 720p rate is $0.125 per output second, so the four rungs are $0.0375, $0.125, $0.1875 and $0.375 per second. A rule of thumb follows: a 4K clip costs as much as three 720p clips of the same length, and a 360p draft costs less than a third of a 720p one.

The ladder

Sume bills the provider list times 1.25 for each output second, as a function of resolution. The provider list in Sume's price table is $0.03, $0.10, $0.15 and $0.30, which is where the ratios come from. The 8-second and 10-second columns show the cost of a typical clip and of the longest single job.

What the rungs mean at Google

Google's Omni page lists the outputs as 360p, 720p (the default), 1080p (upscaled) and 4K (upscaled), and its pricing page states a token conversion only for 720p. So the ladder above is Sume's own price structure, not a copy of a Google price list. The word 'upscaled' matters for choosing a rung: if the 1080p and 4K files are produced by upscaling, the extra cost buys more pixels, not new detail, and a short test clip tells you whether it is worth it for your content.

How to spend the ladder

Draft at 360p. A 10-second draft is $0.375, so ten prompt variants cost $3.75. Pick the best two, re-render those at 720p for $1.25 each, and render the winner at 1080p for $1.875 or 4K for $3.75. Compare that with rendering all ten at 1080p, which is $18.75.

Do the same arithmetic on your own prompts. The table is a rate card, and usage.cost on each completed job is the number your invoice will use.

• Draft: 360p, 0.3x.
• Review and social: 720p, 1x.
• Ads and web: 1080p, 1.5x.
• Large screens: 4K, 3x.

Choosing a rung by what the clip is for

A 360p clip is for judging motion and framing, not for publishing. A 720p clip is fine on a phone feed. A 1080p clip is the usual choice for an ad that will be reviewed on a laptop. A 4K clip only pays when it will be shown on a large display, and because Google describes both 1080p and 4K as upscaled, check one sample at that rung before you commit a batch to it.

If all your clips are vertical for social feeds, 720p at 9:16 is usually the first rung to test; if a platform recompresses your upload anyway, a higher rung can be wasted money.

One more check

A replay with the same Idempotency-Key returns the original job, according to the Sume docs. If you change the resolution, change the key; otherwise the replay returns the earlier job and you will not get the new rung.