Two long lists that disagree about what a language is

Synthesia publishes 143 named rows mixing languages and dialects in one table. Fliki publishes 91 languages and 117 dialects in two separately counted sections. The two disagree about the unit rather than about the size. As of 2026-09-22.

Inclusion rule. The two longest named spoken-language lists in this register. Rows compare what each list counts rather than how long it is. Order. Entry count first, then the unit, then the figures around each list.

1A hundred and forty-three of what

One table counts languages and named regional variants in a single sequence, so its total is not a count of languages. The other splits the two and totals each, so its figures are.

The difference is not academic. A producer converting a headline into a market list needs to know how much of it is variants of languages already present, and only one of these two disclosures answers that without reading every row.

2One language, fourteen rows

Inside the mixed table, Arabic appears as a base entry plus thirteen named regional variants from Bahrain to Yemen. That is nearly a tenth of the table spent on one language.

It makes the table more useful for a Gulf buyer and the total less useful for everybody. Both effects are real, which is why the register records the table as a named list and does not treat its length as comparable with a count of languages.

3Both marketing figures miss their own lists

One vendor prints two different figures above its table, twenty apart, both higher than the row count. The other prints figures below its own totals in both units, understating the list by eleven languages and seventeen dialects.

Overstating and understating are different failures and the register handles them the same way: quote the list, record the headline beside it. Neither pair of figures can be reconciled with the documentation without the vendor saying more.

4Sources

Counted from the Synthesia supported languages table and the Fliki languages and dialects page. Both rows are kept in the register under one rule, set out on how counted; the column is described on dialogue language.

5Both of them against the rest of the column

Inclusion rule. Tools with a public English site selling AI video generation or AI video dubbing. Site languages count hreflang alternates served from the host being read; x-default and alternates pointing at another host are not counted, for every vendor alike. Order. Alphabetical by tool name.

Other pairings that turn on the same kind of difference: PixVerse and Vidu, HeyGen and Rask AI. The full set of pairings is here; single cells are indexed at entries.

• Voice languages
  A table headed “All languages and dialects supported for Synthesia translation feature”, running 143 named rows from Afrikaans to Zulunamed list publishedSynthesia Docs, supported languages / recorded 2026-09-22
• Voice languages
  A page headed “Languages and Dialects” with two named sections, “All languages (91)” and “All dialects (117)”named list published with its own totalsFliki, languages and dialects / recorded 2026-09-22
• Voice languages
  The translator feature page says “80+ languages” and “100+ dialects”, both below the totals printed on the list pagemarketing count under the list countFliki, translator feature page / recorded 2026-09-22