15-second AI video ad: which models make it in a single clip on Sume

Every Sume video model except Gemini Omni Flash 1.1 reaches 15 seconds in one request, and Seedance 2.5 and Wan 3.0 go to 30. For a 15-second social ad the useful question is not length but what the ad needs around it: spoken audio, a reference product photo, a vertical frame or a final 1080p.

Which models, and what each adds

Gemini Omni Flash is the exception: its ceiling is 10 seconds, so a 15-second ad there means two clips joined.

A decision in three questions

Answer these in order and one or two models remain.

• Is your product a photo you must keep exact? Use image-to-video with the photo as the first frame on Seedance, Wan, Kling or MiniMax.
• Do you need the model's own speech or ambience? Use MiniMax or Gemini Omni Flash (always on), or a toggle row with generate_audio true.
• Is the final 1080p? Kling 3, Seedance, Wan and MiniMax H3 Max reach it; MiniMax H3 stops at native 768p unless you pay for an upscale.

Make it cheaper to iterate

Draft the concept at the lowest resolution the model supports, pick the take, and only then render the final. Each rerun is a new take because the v1 models do not accept a seed.