AreaAudit

Which languages each generator says it supports, and where

The variant question, and who it costs money

For most languages a name is enough. For a handful it is not: the national forms differ enough that an audience rejects the wrong one, and almost no published language list says which form it means. As of 2026-09-12.

Where the variant question costs money, layer by layerA marketing page in the wrong variant reads as slightly foreign. A subtitle in the wrong variant uses unfamiliar vocabulary. A dubbed voice in the wrong variant is rejected within a line, and it is the layer where nothing published states which variant is in use.Cheapest layer, best documentedA marketing pageSpelling and idiom. Sometimes answered by a region subtag.A subtitleVocabulary a market may not use. Rewritable for the cost of afile.A dubbed voiceRejected inside a sentence. No voice figure anywhere names avariant.Most expensive layer, no disclosure at all
Fig. 1 The cost rises down the layers and the disclosure thins out at the same rate.
Where the variant question changes what an audience gets. Recorded 2026-09-12.
LayerVariant mattersUsually stated
A marketing pageA little, in spelling and idiomSometimes, as a region subtag
A subtitleYes, in vocabularyNo
A dubbed voiceDecisively, within a sentenceNo
A generated documentRarelyNo

Inclusion rule. The layers a national variant affects, with whether any vendor page tends to state which variant is in use. Languages with one national standard are unaffected. Order. Cheapest layer first.

1The split is audible and commercial

A viewer in Mexico hearing Castilian Spanish, or one in Lisbon hearing Brazilian Portuguese, notices inside a line. Distributors have commissioned separate versions for decades, which is the strongest evidence that the difference is not academic.

The same is true of Canadian against European French and of the two written forms of Chinese. Those four cases cover a very large share of the non-English audience for short video, which is why the question keeps arriving.

2A code can answer it and a count cannot

A language code with a country attached names the variant. A code without one does not, and no count of any size does. So the only published variant information in this category comes from markup rather than from marketing.

That is an odd place for it to live, because markup describes the vendor own pages and the variant matters most in the output. The layer with the evidence is not the layer with the question.

3Where the answers point, they point the same way

Where a variant is named in a declared set, it is usually the smaller of the two candidate markets. Brazil is the exception and is named more than once; Spain and France are named where the larger audiences elsewhere are not.

That is worth knowing before reading any set: a tagged code is a specific claim about one market, and the specificity can rule a reader out as easily as it rules them in.

4What a buyer can do about it

Ask directly, and ask about voices rather than pages. The question has a short answer at every vendor and it is not on any page, so it is the single most efficient thing to put in a first email.

For a subtitled release the question is cheaper: a subtitle can be rewritten per market for the cost of a file. For a dubbed one it decides which voices exist and therefore whether the release is possible at all.

An explainer about method and structure, not a count. No vendor number appears here that is not also in the register. The sourced material is on the register. Related: Blanks as evidence, Two-sided lists.