720p vs 1080p AI Video: What the Extra Pixels Cost, Model by Model

The 1080p switch looks like one setting but it is priced four different ways across Sume's catalog. A model with flat pricing makes 1080p free, a tiered model makes it a step, and a token-priced model makes it a function of pixel count. Knowing which you have changes where you should spend.

The steps

Sume rates are provider list times 1.25. Seedance rows are computed from token counts (width times height times seconds times 24, divided by 1,024) at 16:9, so they say about.

Why Seedance jumps

A 1080p 16:9 frame is 1920 by 1080, which is 2.25 times the pixels of 1280 by 720. For Seedance 2.5 the 1080p provider rate is also higher, $0.0234 per 1,000 tokens against $0.0214 at 720p, so the step is about 2.5 times. MiniMax H3 Max is different: its 1080p is a latent refinement from native 768p, per Sume's docs, so you are paying for refinement rather than a native render.

For reference, Google's pricing page lists Veo 3.1 Standard at $0.40 per second for both 720p and 1080p, and $0.60 at 4K. Flat 720p to 1080p pricing exists elsewhere too.

How to choose

If the platform will recompress to 720p or 1080p anyway, spend on 720p and put the budget into more takes. If you need a real 1080p master, the cheapest route is Kling 3, where the step is free, or Omni at 1.5 times. Seedance 2.5 at 1080p is the premium option and deserves a single hero shot, not a batch.

Check each model's tiers in GET /v1/videos/models and pass resolution explicitly. See the five questions before pinning a model.