A localization can begin as a compromise and end as the face of the thing it compromised.
The record is supposed to prevent that.
The source-language version came first. It has the earlier date, the earlier text, the cleaner claim on origin.
A later regional version has translation and approval.
And whatever commercial weather it had to pass through on the way.
Then a crossover arrives.
It needs a character people recognize now.
Not a production history. Not a source tree. Not an adjudication of what came first.
A usable name. A usable look. A usable set of permissions shared by the host game, the rights holder, and the market receiving the event.
Keep licensing. Keep translating. Keep making a local exception familiar enough that it stops feeling like one.
Recognition is not canon, but it can become infrastructure.
The two systems that produce these collaborations begin in very different places.
The media-mix model is decades old.
A character travels across television, film, games, manga, toys, and whatever else will hold it.
Each appearance is planned to make the property more legible in the next room.
That’s franchise construction.
Everything is built to point at everything else.
A live-service crossover event belongs to another lineage entirely.
It arrives inside a game already running.
An operations calendar needs occasions.
It licenses a narrow asset set for a dated window. It does not build a new work from the ground up. An expansion adds to a franchise; an event adds to a calendar.
They can use the same character without asking that character to do remotely the same job.
Pokémon makes the difference unusually visible, because its chain has names attached.
The Pokémon Company exists as a joint venture of Nintendo, Creatures, and Game Freak.
Nintendo holds under a third of it.
It manages brand licensing, production, and marketing. Game Freak develops the mainline games. The anime has been produced since the beginning by a studio outside that structure entirely, as a commissioned partner.
A separate international arm handles the same functions in other regions.
And the card game.
Five parties stand behind what casually looks like one cheerful mouse on a banner.
That isn’t pedantry.
It’s the difference between the rights holder, the licensing authority and the developer.
And the studio whose images the public most likely has in mind.
A crossover doesn’t license the anime just because the animation made the character famous.
It traverses a chain.
The chain decides what may appear.
The chain is historical as well as structural.
A long-running mobile game can keep the same name on its store listing while the studio of record is spun out of its publisher, established as a new company, and acquired back within weeks.
The name on the game stays continuous. The developer does not.
Whoever negotiates a crossover two years later negotiates with the current arrangement.
Not the one that made the game.
Release variation makes a similar inconvenience visible inside the work.
Post-gold localization adapts a game after the source-language version is finished.
Sim-ship builds the regional releases in parallel.
The first structurally produces a gap between territories. The second spends real money to erase one.
That gap isn’t just a waiting audience.
It’s two different production states.
Sometimes two different technical ancestors.
A localized build can be reconstructed from a later revision of the source rather than translated from the original.
Because the original code was no longer usable.
When that happens, the regional release isn’t the source game with new words pasted over it.
It has inherited a different past.
Then a partner needs a character, and none of this survives the meeting.
From inside a crossover menu, every version looks equally simple.
One character. One approved representation. One asset package, cleared for one window.
The event needs the version that can be named, displayed and sold.
And recognized in the territory where it appears.
That’s an inference from the structure of a fixed-window licence. It isn’t a claim that the localized version was secretly the artistic original all along.
But the selection has consequences, and they run in one direction.
The chosen version becomes the one a new audience meets first. It becomes the reference image, the voice, the name in the search box.
There’s a real cost to treating that as the franchise version.
It flattens the translation labour, the earlier releases, the abandoned names, the regional objections.
And the source build’s own claim on the work.
It converts an answer to a market into an answer to history.
The opposite position charges its own cost, and it’s easy to underrate.
Insisting on an unaltered origin asks a local audience to recognize a character by a name, a rule, or a build it never received.
That protects a lineage by making legibility somebody else’s problem.
Usually the audience’s.
The commercial case here isn’t shallow.
A crossover is shared public language operating under a limited licence.
Shared language has to be shared to work.
The regional case isn’t parochial either.
A localized build can carry a distinct codebase and a distinct compliance history.
And years of ordinary use by people who never saw another version.
The source build preserves how a property began.
The localized build preserves how a market actually met it.
The crossover makes one of those histories newly visible.
It settles nothing about which of them deserves the throne.
So canon turns out to be the wrong thing to be arguing about.
Which version has a collaboration made it possible to recognize?