Shortest AI video clip each Sume model accepts, and its price

Every video model on Sume has a floor on duration. The floor decides the cheapest request you can make on that model, which matters when you want to check a prompt, a reference image or an API integration without paying for a full-length clip.

Limits are from the Video Router docs and the Avatar Video docs. Prices are Sume billed rates read on 2026-10-03.

What is the minimum request on each model?

Wan 3.0 accepts 2 to 30 seconds, the lowest floor in the table. Gemini Omni Flash 1.1 accepts 3 to 10 seconds. Seedance 2.5 accepts 4 to 30 seconds, MiniMax H3 and H3 Max 5 to 15 seconds. The video-to-video models bill by the source clip's length, so their minimum is the shortest source you can send.

Why test at the minimum?

A 2-second Wan 3.0 clip at 480p is $0.125, so an integration test of 100 requests costs $12.50. A 4-second Seedance 2.5 clip at 480p is $1.0747, about 9 times that. The video-to-video models have the dearest cheapest-tier minimums: $1.59 for Genjutsu at 480p and $1.875 for H3 Max Recast at 768p.

Minimum clips prove that a request validates, that the response shape is what your code expects and that the charge matches the table. They do not predict how a longer clip will look, because motion and consistency change with length.