Which video model fits a 6, 15 or 30 second ad slot on Sume?

Pick the model whose duration range contains your slot, because limits are per model. In the Sume Video Router, seedance-2.5 takes 4 to 30 seconds and wan-3.0 takes 2 to 30, gemini-omni-flash-1.1 takes 3 to 10, minimax-h3 takes 5 to 15, and every other catalog model has a 15 second limit. A 6 second bumper fits nearly all of them, a 30 second spot needs one of the long ones or a timeline of shorter clips.

Duration limits by model

Read capabilities from GET /v1/video-router/models before you submit, because each model can have a different envelope and a model is listed only when its provider is configured.

Let Sume choose

With POST /v1/videos and model: "sume/auto", the create controls default to 720p and 8 seconds, with 3 to 10 second clips at 16:9 or 9:16. That covers short bumpers and hooks. The Video Router docs recommend Auto for new work and the explicit catalog when you must pin a model.

Auto is the right default when the slot is short and the model does not matter. It stops being the right default when a client has approved a specific look, or when the slot is longer than 10 seconds, because then the explicit catalog is the only way to be sure of the limit that applies.

A 30 second spot, two ways

One call to a long model gives one continuous clip. The alternative is several shorter clips composed with timeline-1.0, which accepts 1 to 200 video[] slots, up to 1800 seconds of audio, with transitions of up to one second. The timeline path costs $0.10 per started output minute on top of the generation, and it gives you control per cut, which suits ad variants where only the hook changes.

Request sample

The docs example for pinning a model is a vertical UGC style clip at 720p and 12 seconds on seedance-2.5, sent to POST /v1/video-router/generate with mode: "async" and an Idempotency-Key. Poll the job and read the result URL when the status is ready.

Whichever route you pick, bill-aware habits apply: the Video Router bills list price times 1.25 for each model, per the docs, and each generation is an async job with an Idempotency-Key. For a variant set, hold the slot length constant across models only if you are comparing models. If you are comparing hooks, hold the model constant and vary the prompt.