The same six shapes, opposite defaults
Both pages publish 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16. One adds adaptive as the default, letting the model choose. The other states that the field is required and that adaptive is not offered. Identical lists, opposite behaviour when nobody sets the field. As of 2026-09-22.
| What the pages state | Wan 3.0 | MiniMax |
|---|---|---|
| The six named values | Identical on both pages | Identical on both pages |
| Default | adaptive: the model chooses | None; the field is required |
| Risk in an unattended batch | Mixed shapes across one campaign | A rejected request |
Inclusion rule. Shape statements from each page, including the default each names. The risk column states the consequence for a booked placement rather than a vendor claim. Order. The shared list first, then the behaviour that differs.
1A failed request is cheaper than a wrong shape
The required field turns an omission into an error at submission time. The adaptive default turns the same omission into a batch of usable files in shapes nobody booked, discovered at upload.
For advertising the first failure mode is plainly preferable. It costs a retry; the second costs the generation budget and the schedule, because the pieces have to be made again after the problem is understood.
2Nothing about the lists explains the difference
Six values, both including ultrawide and square, both applying to every tier the same page publishes. If the register carried only the list, these two rows would be interchangeable, and in production they are not.
That is why the default sits in the cell rather than in a footnote. A shape column recording only which shapes exist would describe the easier half of the question and leave the half that causes rework unrecorded.
3Both rows carry the fix in one line
Set the ratio on every request. On the adaptive row that overrides the default; on the required row it is the only way to submit at all. The same discipline works on both, which is worth saying because most of this column has no documented default to override.
What differs is what happens when the discipline lapses. One row tells you immediately and the other tells you after the media is booked, and that is the whole content of this comparison.
4The rest of the column behind these two
| Generator | Aspect ratio |
|---|---|
| Amazon Nova Reel | Not stated |
| CogVideoX | Not stated |
| Google Veo | 16:9 and 9:16 |
| Grok Imagine | Seven, 16:9 the default |
| Hedra | Seven, including 9:21 and 21:9 |
| HeyGen | 16:9 or 9:16 |
| Kling AI | Not stated |
| LTX-Video | Not stated |
| MiniMax | Six, required for text-to-video |
| Pika | Not stated |
| PixVerse | Eight on version 6 |
| Runway | Seven pixel pairs |
| SceneMixer | 9:16 and 16:9 |
| Vidu | Not stated |
| Wan 3.0 | Six, adaptive by default |
Inclusion rule. Every generator this register carries, on the one column this pair disagrees about. Rows that state nothing keep their place and are marked hollow. Order. Alphabetical by generator name.
Why these two. Two pages offering the same six values behave differently in production because one proceeds on its own judgement and the other stops. That is the clearest demonstration on these sheets that a default is a specification.
- Aspect ratios offeredadaptive, the default, then 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16
- Aspect ratios; for image-to-video the ratio follows the input image21:9, 16:9, 4:3, 1:1, 3:4 and 9:16, required and never adaptive for text-to-video
5Sources
Read from the Wan3.0 API reference and the MiniMax video task reference, both checked 2026-09-22. The column is described on aspect ratio, and each entry has its own cell page at Wan 3.0 and MiniMax, and the rule that keeps surfaces and generators apart is on method. Nearby pairs: Grok Imagine and Pika, LTX-Video and Wan 3.0.