The questions a specification cannot answer
A register of published figures settles whether a file can fill a slot. It cannot settle whether the ad will be approved, whether it will work, or what it will really cost to make, and pretending otherwise is how spec sheets become misleading. As of 2026-09-12.
| Question | Settled by published figures | Decided by |
|---|---|---|
| Does the file fit the slot | Yes | Arithmetic on two published numbers |
| Will the platform run it | No | A review process no page describes |
| How many attempts will it take | No | Nobody publishes this |
| Will it perform | No | An auction and an audience |
Inclusion rule. Questions a buyer commonly brings to a specification comparison, separated by whether published figures can answer them. Order. The answerable question first.
1The attempt count is the missing budget line
A rate per second multiplied by a length gives a floor, not a forecast. What decides a real budget is how many generations it takes to get a usable one, and no page in this genre publishes anything about it.
So an estimate built purely from published rates will be low, and the multiple is unknown rather than small. A plan should carry a contingency whose size comes from experience rather than from a table.
2Approval is a different question from fit
Whether a platform will run generated creative, and what it requires to be disclosed, is decided by policy and by people. A file can satisfy every published specification and still be refused, and the specification pages do not speak to that.
Keeping the two apart is a discipline rather than a limitation. A register that mixed fit and approval would be guessing about the second while being precise about the first.
3What the narrow question is worth
It decides the production plan. Whether a placement is one generation or an assembly changes the schedule, the roles involved and where the creative risk sits, and it can be worked out before anyone opens an editor.
That is a small answer delivered reliably, which is more useful than a large answer delivered on a guess. The rest of the questions are real and they belong to people with access to systems no published page describes.
A craft note, not a spec. Nothing here is taken from a platform or a vendor, and nothing here should be quoted as a requirement. The sourced material is on published limits. Related: One master, many cuts, Failing at upload.
4Making the ad itself
For ad work around “What a sheet cannot settle”, SceneMixer's pages explain how its ad projects are produced:
An AI short drama is a serialized show of 60–120 second vertical episodes whose cast, sets and footage are generated by AI. Making one takes six steps: analyze the script into episodes, lock a consistent cast with reference art, localize for your market, write and edit storyboards, generate each 4–15 second shot, then composite and publish.
SceneMixer, How to Make an AI Short Drama, Step by Step (read 2026-10-13)
The Seedance 2.0 tiers render segments up to 15 seconds; use a Seedance 2.5 or Wan tier for 30-second takes.
SceneMixer, Seedance 2.0 Fast — “Does Fast render 30-second takes?” (read 2026-10-13)