Seedance 2.5 vs 2.0: what changes in Sume's catalog rows

What is different between Seedance 2.5 and 2.0 on Sume? Length is the big one: seedance-2.5 runs 4 to 30 seconds, while seedance-2, seedance-2-fast and seedance-2-mini run 4 to 15. The list rate also moves, from $0.014 per 1,000 video tokens for Seedance 2 to $0.0214 for 2.5. The older ids stay routable by explicit id.

What the vendor adds

The Seed announcement describes up to 30 seconds per generation with multi-round extension, up to 30 images, 10 videos and 10 audio clips per pass, and timestamp-level, green-screen and camera perspective editing. I did not read a Seedance 2.0 vendor page today, so this post compares only the Sume rows against the 2.5 announcement.

Choosing between them

Choose seedance-2.5 when a clip must exceed 15 seconds or when you want the larger reference pass. Choose seedance-2-fast or seedance-2-mini when volume and price matter more. Sume bills list times 1.25, rounded up to cents, and token counts depend on the clip, so run one short test to see the real charge rather than multiplying by guess.

Read pricing_skus from `GET /v1/videos/models` for the current rates, since this table is a snapshot.

Same shape, longer reach

All four Seedance rows share the same reference types and the same 480p, 720p and 1080p tiers in Sume's catalog, plus audio. So an existing Seedance 2.0 integration should work against 2.5 by changing the model id and, if you want longer clips, the duration. Check the aspect ratio list, since the catalog lists 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16 for Seedance 2.5, and keep a test clip of each shape you ship.

Check the request fields in the video models docs before you reuse an old payload.

Prices on the 2.0 rows are lower per token, but a longer clip costs more tokens, so compare cost per finished second on your own content. A 4 second test at 480p on each id is a cheap way to see the gap before you commit a batch, and the result tells you more than any table.