Wan 3.0 vs Seedance 2.5: 30-second clips, inputs and price

Wan 3.0 and Seedance 2.5 are the two models on Sume's POST /v1/videos that make a single clip of up to 30 seconds, and both take a prompt, a first and last frame, image, video and audio references, and optional audio. What separates them is the meter and the reference caps: Wan bills a flat rate per second at each resolution, Seedance bills per 1,000 video tokens, so frame size and aspect ratio move its price; Wan caps each reference type, while Seedance lists no per-type caps. Wan also starts at 2 seconds against Seedance's 4, and Seedance adds a 21:9 frame.

The limits are those of wan-3.0 and seedance-2.5, from Video generation, Video Router, the Sume API reference, and the catalog and pricing code, read on 2026-09-28. Nothing here ranks their output.

How do Wan 3.0 and Seedance 2.5 differ?

Resolutions, inputs, audio and the 30-second ceiling match, so only the differing rows are listed; Wan 3.0 API and Seedance 2.5 API carry the rest.

What does a 30-second clip cost on each?

Wan's price depends only on resolution and length. Seedance counts video tokens, width × height × seconds × 24 fps ÷ 1,024, so size and aspect ratio both move it; audio changes neither in current code. Always send resolution to wan-3.0: without it, Sume reserves at the 1080p rate.

The amounts below are reserved on submit, plus a 5.5% agent fee by default.

How many references can each one take?

Wan 3.0 lists per-type caps, and current code refuses a request with more references of a type than its count cap:

• Wan 3.0: up to 10 images, up to 5 videos (15 seconds in total, each at 16 fps or more), and up to 5 audio clips (15 seconds in total).
• Seedance 2.5: image, video and audio references, within the /v1/videos limit of 12 input_references per request; its catalog entry lists no per-type caps.
• Both: a request that carries frames and references is treated as image-to-video, and the references are dropped.

Is Wan 2.7 or Seedance 2.0 on Sume too?

Seedance 2.0 is, as seedance-2, with the same inputs as 2.5 but a 15-second cap. Wan 2.7 is not: wan-3.0 is the only Wan id in the catalog.

Which should I pick?

• Pick Wan 3.0 for clips shorter than 4 seconds, or a price that ignores frame shape.
• Pick Seedance 2.5 for a 21:9 frame.
• Either one for 16 to 30 seconds in one clip; beyond 30 seconds, chain clips.
• Price the exact request you plan to send: for Seedance, resolution and aspect ratio both change the token count.