Does sume/auto pick Seedance? Sume does not say which model ran

You cannot tell, and Sume says you should not try. With model: "sume/auto" the poll response reports sume/auto, and the docs say the family that served the request is never disclosed. If you must have Seedance, send seedance-2.5 or seedance-2 as the model.

The design is intentional: it lets Sume change the underlying choice as models improve without breaking your integration.

What does sume/auto promise?

The docs say resolution is a pure function of the normalized request and the catalog version, so an idempotent replay prices and routes the same way. They also warn against inferring the model from any visible trait of the output.

The consequence is that two identical requests made months apart under auto may be served differently if the catalog version changed. Within one catalog version the docs say the result is deterministic for the same normalized request, which is what makes idempotent replays price and route identically.

What are its limits?

Auto create controls default to 720p and 8 seconds, with 3 to 10 second clips at 16:9 or 9:16, per the Video Router docs. Seedance 2.5 alone goes to 30 seconds, so a long clip needs a pin.

When is auto the right choice?

When you care about a plain clip and not which engine renders it: a short vertical product loop, a draft. For an A/B between models, a pinned id is the only way to keep the experiment honest, because auto may change as the catalog changes.

A decent rule: use auto for drafts and for content where the engine does not matter, and pin for anything where look, length, references or audio behavior is part of the brief. If a client approved a look from Seedance output, pinning protects that approval.

Pinning also gives you honest numbers. When you report cost per clip or turnaround time to your team, you want to know which model they describe. Under auto the figure blends whatever ran.

How do I pin?

Replace the model string and keep everything else. Check the ratio and duration against GET /v1/videos/models first, because a request that was valid under auto may fall outside the pinned model's list, for example a 3-second clip on a model whose minimum is 4. See auto's duration limit.

Also note what pinning does not give you: it does not unlock a model your workspace cannot reach. Check the catalog list for the ids available to you before you hard-code one.