Gemini Omni 1.1 Flash 1080p and 4K price: upscales on Sume

Is Gemini Omni 1.1 Flash really 4K? Google's docs describe 720p as the default and 1080p and 4K as upscaling options, with a 360p draft mode. So a 4K file is an upscaled 720p-class render. On Sume, gemini-omni-flash-1.1 lists 360p, 720p, 1080p and 4K, and a 10 second clip costs from $0.38 to $3.75 depending on the tier.

What Google says

The Gemini API Omni docs list 360p, 720p default, 1080p upscaled and 4K upscaled, with 16:9 and 9:16 aspect ratios. Google's launch post says draft mode at 360p is up to 60% faster and at a third of the cost.

On Google's own pricing page, Omni video output is $17.50 per million tokens, about $0.10 per second at 720p.

Sume's per-second rates

Sume's catalog list rates are $0.03, $0.10, $0.15 and $0.30 per second for 360p, 720p, 1080p and 4K, and Sume bills list times 1.25 rounded up to cents. The 720p list rate matches Google's roughly $0.10 per second.

Buying advice

Draft at 360p to check the idea, render the keeper at 720p, and request 1080p or 4K only for delivery. If the final goes to a platform that will recompress it, a 720p render may look identical.

Prices change; read pricing_skus from GET /v1/videos/models for the live numbers.

What the table does not include

The numbers above assume the default video output with no extras. Sume applies list times 1.25 per clip, rounded up to cents, and the list rates here come from Sume's catalog pricing tables. Native audio is part of the model, so there is no audio toggle to price. If your clip is shorter than 10 seconds, scale linearly by seconds, then round up to the cent.

Drafting cheaply

Google's post ties draft mode to 360p and says it runs up to 60% faster at a third of the cost. Sume's 360p list rate of $0.03 per second against $0.10 at 720p follows the same one-third shape, which makes 360p a sensible place to test prompts before spending on a keeper.

One caution on reading Google's tiers: the docs label 1080p and 4K as upscaled, so choose them for delivery formats rather than for extra detail in the generation itself.