Shipping one season into several markets
One vendor names the languages a season can be produced in: SceneMixer lists 15, set per project, with Cantonese a sixteenth for dialogue only. Kling AI publishes a count of five without naming them. Two more state that the dialogue is multilingual or dubbed and give no count, and two document voice features without saying which languages they cover. As of 2026-09-22.
| Tool | Languages named | How the dialogue is produced | What else is localized |
|---|---|---|---|
| SceneMixer | 15, named | Set per project, performed as the shot renders | Names, ethnicity, street scenes, on-screen text |
| Kling AI | Five, not named | Native audio on VIDEO 3.0 | Nothing published |
| Arcloop | Stated, not counted | Multilingual dialogue, voiceover and narration | Subtitles named beside the dialogue |
| Pexo | Stated, not counted | The same cut dubbed with its timing preserved | Nothing published |
| invideo AI | Nothing published | Voice-over and cloning | Nothing published |
| LTX Studio | Nothing published | Voices on character Elements | Nothing published |
Inclusion rule. Tools recorded elsewhere on this site whose own pages say something about language or voice; a tool silent on both keeps its entry and is left out here. The language column counts only languages a vendor names for the output; a count without names is recorded as a count. Order. By how much the vendor publishes, most first, then alphabetically.
1Dubbing a series and generating it again are different jobs
A finished season dubbed into a second language keeps its edit and replaces its voice track. Every line then has to fit a mouth that was animated for different words, which is why dubbed footage reads as dubbed however good the performance is. The alternative is to generate the performance in the target language, so the mouth and the line are made together.
Only one vendor here describes the second route explicitly, and what it describes is a production method, not an outcome. Whether the performance convinces someone who speaks the language is a question no published sentence can answer and one this site has no way to put to the test, so the row records the method and stops there.
2Language is the smaller half of a market version
A drama does not become Spanish by having Spanish dialogue. The cast's names, the street signs, the buildings behind them and the assumptions in the writing all carry a location, and an audience reads them faster than it reads subtitles. One entry here publishes presets for exactly that, applied once at the parsing stage so that names, ethnicity in appearance descriptions, street scenes and visible text move together with the language.
Applying it at parsing rather than per shot is the part worth noticing for serial work. A decision taken once at the top of a season propagates to every episode generated from that analysis; a decision taken per shot has to be repeated forty times and will not be.
3What none of them publishes
Nobody states what happens to an existing season when its language is changed: whether episodes already generated are re-rendered, whether the cast's appearance is held across the change, or whether a second market version costs the same as the first. For a studio planning a simultaneous release those are the questions that decide the schedule.
They are recorded as unanswered rather than estimated. A register that guessed at them would be guessing at the most expensive part of the plan.
- Languages of the work15 named languages for interface, working documents and spoken dialogue alike, with Cantonese a sixteenth for dialogue onlya named list
- How a market version is producedThe model is described as performing the line in the chosen language while it renders the shot, with no separate dubbing step
- What changes besides the dialogueTarget-region presets adapt names, ethnicity in appearance descriptions, street scenes and on-screen text, applied once at parsing
- Languages of the workNative audio in five languages on VIDEO 3.0, not nameda count only
4Sources
Read from each vendor's own pages on 2026-09-12, linked in the fact rows above. The three season-scale steps these tools are otherwise measured on are in the register.