Which entries mark the models they made themselves
Almost nobody marks the difference. Several entries put their own models and other makers' on one table in the same type, with a reader expected to already know which maker each name belongs to. One acknowledges the mixture in its plan wording without marking any row. From 4 source pages, read 2026-09-22; no maker pays to be listed.
| Entry | What is published | Detail |
|---|---|---|
| Luma | The plan wording mentions other makers | Individual rows carry no marking |
| Adobe Firefly | Partner models are named per tier | Which is the nearest thing to marking |
| PixVerse | A house line listed with licensed models | No marking distinguishes them |
| Runway | House and licensed rows, same conversion table | No marking on any row |
| invideo AI | Every line is another maker's | So nothing needs distinguishing |
Inclusion rule. An entry appears when its own public pages name models from more than one maker. Marking counts when the page itself distinguishes them, not when a reader can. Order. Entries acknowledging the mixture first, then the rest.
1Two failure modes in the same typeface
A house line goes when the seller decides. A licensed line goes when somebody else does and arrives as an edit. Presenting them identically hides which kind of ending a reader is exposed to.
Marking the licensed rows would change nothing about the risk and everything about whether a reader can see it. Three entries reached the same unmarked arrangement without coordinating.
2Tier gating is accidental marking
Where partner models are named per plan tier, a reader learns which models are partners as a side effect of learning which plan reaches them. That is more information than an unmarked table gives.
It also creates a second way to lose access, because a tier can move without anything being retired. Disclosure and exposure arrive together.
3An aggregator needs no marking
An entry with no models of its own has nothing to distinguish, so its list is honest by construction. It also has no floor under it: if every licensed line went, nothing would remain.
The page setting the two layers of aggregation against each other shows what that difference costs in the worst case.
- How the plan describes the menu it sellsPlans are described as covering Luma models alongside models from other makers
4Production that outlasts a model
Model end dates touch “House against licensed”; SceneMixer's guides describe how its pipeline treats the models it runs:
Dubbing replaces the audio of a finished clip; the picture was generated for a different line, so mouth shapes, pauses and shot length all belong to the original language. Native-language generation gives the video model the line in the target language before it renders the shot, so the performance, the pacing and the cut points are built for that language.
SceneMixer, Native-Language Dialogue vs. Dubbing in AI Video (2026): Why the Cast Should Speak the Line (read 2026-10-13)
Typically 60–120 seconds, always 9:16 vertical, ending on a hook. SceneMixer's episode outlines are generated with that pacing, and setting a project to 9:16 makes every segment vertical (the default is 16:9).
SceneMixer, How to Make ReelShort-Style Dramas with AI — “How long is a ReelShort-style episode?” (read 2026-10-13)
5Sources
Dates come from the company that makes the model or the platform that serves it; menus come from the app, and both are linked on the calendar. Read 2026-09-22. What counts as a date is on sourcing. Nearby: Downloadable weights, Tier and model, Who publishes a date.