Two routes to one model, documented and not
Both entries reach the same models two ways. One publishes a marking statement for each route and the two disagree. The other publishes one for its plans and does not repeat it on its interface pages. As of 2026-09-22.
| On this question | Wan | Runway |
|---|---|---|
| The consumer route | A console that always stamps | Plans, with watermark-free output |
| The interface route | A boolean field, defaulting to false | API pricing, which omits the line |
| Do the two agree | No, and both are published | Unknown, one is silent |
| What decides the file | The door the request came through | Not established |
| Reason published | Compliance, for the console | None |
Inclusion rule. One row per question on which these two entries can be compared from their own published pages. Questions neither vendor addresses are left out rather than filled with a dash. Order. Fixed order: the mechanism first, then who holds it, then what it leaves behind.
| Generator | Paid tier | Document read |
|---|---|---|
| Runway | Watermark-free on paid plans | pricing page |
| Wan | API parameter, default false | video generation guide |
Inclusion rule. The two entries on this page, with the value each one's own pages support in the column they diverge on. Order. Alphabetical by generator name.
1The route can decide the file, and usually nobody says so
Where a company sells the same model through a console and an interface, the door a file came through can decide what it carries. One of these entries documents that outright: the console always stamps and says the stamp cannot be switched off there, while the API publishes a watermark field that defaults to false.
The other publishes a marking line on its consumer plans and does not repeat it on its API pages. The register records the consumer plans, because that is where the statement exists, and does not extend it across a boundary the vendor did not extend it across. So one row has a documented disagreement and the other has an open question.
2What a documented disagreement is worth
It is worth more than agreement would be. A reader integrating the stamping model knows that output from the console and output from the interface are in different states by default, and can plan around it. A reader integrating the silent one knows the consumer plans remove the mark and nothing about the other route.
The vendor with two published statements also gives a reason for one of them, attributing the console restriction to legal compliance requirements. That makes the asymmetry legible rather than arbitrary: the surface a regulator would look at is the one with no switch.
3Where both rows leave a recipient
Neither entry names a provenance standard or a metadata field, so both record not documented for file contents. A clean file from either product carries no indication of which route produced it, whether a decision was taken, or whether anybody knew there was one to take.
On the entry with the documented split, a stamped file at least proves the console route. On the silent one there is no observable difference between the routes at all, which is the weaker position for anybody trying to establish provenance after the fact.
4Sources
- ControlThe console watermark cannot be turned off from the console UI, which the page attributes to legal compliance requirementsno switch on the console surface
- Machine-readable markingNot documented on the vendor's public pages (as of 2026-09-22)not mentioned on the pricing page
Both values above are read from the vendor pages linked in the rows: alibabacloud.com and runway.com, on 2026-09-22. Each entry has its own page, Wan and Runway, and the rule for the column they diverge on is on the paid tier column page. Other pairs: a switch and a benefit, two off-switches.