What 30 Seconds of AI Video Costs at 720p, Model by Model

Per-second prices hide the real spread, because Seedance bills by video tokens and Kling by audio setting. Multiplying everything out to 30 seconds makes the spread visible. All Sume figures are the billable amount: provider list times 1.25, reserved on submit and refunded on failure.

Because most models cap a single clip at 15 seconds, 30 seconds means two clips at the same per-second rate. Cost per second does not change with the split.

The table

Resolutions are 720p unless the model has no 720p tier: MiniMax H3 and H3 Max are first-class at 768p, and Grok Imagine tops out at 720p.

How the Seedance rows are computed

Seedance is priced per 1,000 video tokens in Sume's pricing tables, and a 720p 16:9 clip at 24 fps comes to roughly 21,600 tokens per second. That is why the Seedance rows say about. A reference video input raises the billed amount, so a clip with video references costs more than the table shows.

Reference point from Google

Google's Gemini API pricing page lists Veo 3.1 Standard at $0.40 per second at 720p and 1080p, Fast at $0.10, and Lite at $0.05 at 720p. Thirty seconds is therefore $12.00, $3.00 and $1.50 at Google's list price. Those models are not in Sume's catalog today, so they are a yardstick rather than an alternative.

The cheap end is real but narrow. Grok Imagine needs a source image, H3 always renders audio. Use the table to set a draft budget, then spend on the models that fit the shot.