Five bare codes, one of them opened and read
Higgsfield declares five same-host alternates as bare language codes: German, English, Spanish, Japanese and Korean. One of the five was opened, and the Japanese page is written in Japanese throughout rather than swapped from English. As of 2026-09-22.
| Code | Declared on the host | Opened and read |
|---|---|---|
| de | Yes | Not checked |
| en | Yes | Not checked |
| es | Yes | Not checked |
| ja | Yes | Navigation and headings in Japanese, document declared as ja |
| ko | Yes | Not checked |
Inclusion rule. The declared set for one vendor, with the result of a spot check recorded against the single code it was performed on. Order. Alphabetical by code.
1A declaration is cheap, so one was opened
Counting alternates measures markup, not translation. A tag can point at a page that redirects, that has gone stale, or that is the English page with a switcher on it, and the count would look the same in all three cases.
So one code was opened. The Japanese alternate is a Japanese page: the navigation and the headings are in Japanese and the document declares itself as Japanese. That is one check on one language, recorded as exactly that and not generalised to the other four.
2Why the check is not repeated for every code
A spot check that is done for one vendor and not for the others stops being a uniform measurement. The count is what gets compared across the register; the check is a sanity test on the method, and two of them are enough to show the method holds.
The other check in this register is on a different vendor and a different page type, a price page rather than a landing page. Both came back as real translations, which is weak but genuine evidence that declared alternates here are not decorative.
3Five codes and nothing in the other two columns
This vendor publishes no spoken-language list and no statement about the language of anything it generates. So the entire published language position is five site codes, two of which are the most commonly declared codes in the register.
The set is also completely contained in two longer ones, which puts it in the nested chain that runs from two codes up to thirteen. Being inside a larger set is not a weakness; it means the vendor has translated the pages the field translates first.
4Sources
Counted from the markup of higgsfield.ai and checked against its Japanese page. The column itself is described on site language, and every cell Higgsfield fills is on its tool page.
5The same column, tool by tool
| Tool | Site languages |
|---|---|
| CapCut | 27 |
| D-ID | 2 |
| Fliki | None declared |
| Hailuo | None declared |
| HeyGen | None declared |
| Higgsfield | 5 |
| invideo AI | None declared |
| Kling AI | 20 |
| LTX Studio | None declared |
| Pika | None declared |
| PixVerse | 13 |
| Rask AI | One same-host, 15 on subdomains |
| Runway | 5 |
| SceneMixer | 15 |
| Synthesia | 9 |
| VEED | None declared |
| Vidnoz | One same-host, 11 on subdomains |
| Vidu | 9 |
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.
Neighbouring cells: HeyGen: site language and Kling AI: site language. All of them together: the cell index and the register table.
- Site languages5 languages: de, en, es, ja, kodeclared as same-domain alternates
- Spot checkThe Japanese alternate is served as a Japanese page, navigation and headings included, and the document declares lang=“ja”one of the 5
- Voice languagesNo public list (as of 2026-09-12)no list published