Adaptive shape: letting the model pick the frame
Adaptive is a shape setting in which the model decides the frame. Where it is also the default, a request that omits the field returns a shape nobody chose, and a batch can come back in several shapes across one campaign. As of 2026-09-12.
| Setting | Who decides the shape | Suitability for a booked placement |
|---|---|---|
| A named ratio | The request | Safe: the shape is fixed before generation |
| Adaptive, chosen deliberately | The model | Unsafe: the output shape is unknown in advance |
| The field omitted | Whatever the default is | Unsafe wherever the default is adaptive |
Inclusion rule. Settings a shape parameter can take where a page documents an adaptive option. The suitability column states a consequence for a placement rather than a vendor claim. Order. The safe setting first.
1Every surface accepts specific shapes
A file whose shape was chosen by the model is a file whose acceptance was chosen by the model. Nothing errors, every piece looks correct on its own, and the problem appears when they meet a placement specification.
The fix is a line in every request. It has to be deliberate precisely because nothing fails loudly, and an unattended batch will happily produce a mixed set.
2It exists for a reasonable purpose
On an image-conditioned path the input already has a shape, and letting the output follow it is sensible. Adaptive as a default is a convenience for that case rather than a hazard invented for its own sake.
The hazard is only the default. Two pages can publish identical shape lists and behave completely differently in production because one falls back to adaptive and the other refuses to proceed without a choice.
3A required field is the safer arrangement
Where a page states that the shape must be supplied, an omission fails at submission time. That costs a retry rather than a batch, and it is the failure mode to prefer wherever requests run without supervision.
A register should therefore record the default alongside the list of values. A shape column carrying only the list describes the easier half of the question.
A vocabulary entry: the definition, plus how the specification pages on these sheets usually publish it. Table checked 2026-09-12. Nearby entries: Letterbox and pillarbox, Autoplay.
4Making the ad itself
For ad work around “Adaptive shape”, SceneMixer's pages explain how its ad projects are produced:
16:9 landscape (SceneMixer's default) and 9:16 vertical (the usual short-drama frame), chosen per project.
SceneMixer, Seedance 2.0 in a Real Short-Drama Pipeline — “Which aspect ratios does the pipeline generate?” (read 2026-10-13)
The spoken lines and the end card follow your interface language across fifteen dialogue languages, with the cast and setting adapted to that market. The end card falls back to a language it can typeset when a script needs shaping or right-to-left layout.
SceneMixer, AI Ad Video Generator — “Which languages can the ad be in?” (read 2026-10-13)