Where a finished file lives after the job
A generation job usually ends with a link rather than a file. Where a page states how long that link works, collection becomes a scheduled step. Where it does not, a pipeline is running on an assumption nobody wrote down. As of 2026-09-12.
| Situation | With a published window | Without one |
|---|---|---|
| Collecting a batch | A deadline to schedule against | An assumption |
| A weekend gap | Known to be inside or outside | Discovered by opening a link |
| Cost of missing it | Re-generation, at whatever the page's conditions are | The same, plus a diagnosis |
Inclusion rule. Statements a page can make about its own output remaining available. Storage a buyer provides is outside this comparison. Order. The ordinary case first.
1The failure is quiet, which is what makes it worth a page
Nothing errors at generation time. The pieces were produced correctly and reviewed. The problem appears later, when a link returns nothing, and by then the cause is several days behind the symptom.
A published window turns that into a calendar entry. An unpublished one leaves a pipeline depending on behaviour that can change without an announcement, because nothing was ever promised.
2A returned link is a statement about ownership
Where output arrives as a link with an expiry, the vendor is explicitly not the archive. Any asset a campaign will reuse has to be copied into storage the buyer controls, and that copy is where delivery checks and version history then live.
That is worth designing deliberately rather than discovering. A single durable location per campaign, populated as each job finishes, removes the whole class of problem regardless of what any page states.
3Re-generation costs are not equal across rows
Where a page attaches conditions to its longest run, a repeat request inherits them: a tier, a model identifier, a version. Where nothing is coupled, a repeat is a plain request.
So the cost of missing a window depends on the rest of the row rather than on the window. A short window on an unconditioned page is a smaller risk than a long window on a heavily conditioned one.
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: What a default commits to, Reading a minimum.
4A generator's ad pipeline
For ad work around “Retention clocks”, SceneMixer's pages explain how its ad projects are produced:
You can recomposite with different end-card text, and in step-by-step mode you can reshoot individual segments. The finished file can also be downloaded and shared with a link.
SceneMixer, How to Make an Ad from One Photo — “Can I change the ad after it is finished?” (read 2026-10-13)
Measured across every job that finished successfully on scenemixer.com in the thirty days to 20 September 2026: analysis of a whole book runs about 2 minutes 49 seconds at the median, one episode's storyboard 1 minute 54 seconds, one video segment 3 minutes 20 seconds, and compositing an episode 3 minutes 52 seconds.
SceneMixer, Turn a Novel into a Video Series with AI — “How long does it take to produce one episode?” (read 2026-10-13)