AreaAudit

Which languages each generator says it supports, and where

Source language: what a tool can take in

A source language is the language of the material a tool takes in: an uploaded recording, a pasted script, a novel. It is a separate list from the output, and almost nothing in this register publishes it. As of 2026-09-22.

Four kinds of input, and how many are documentedA recording to be dubbed has to be recognised first, and one vendor publishes that list. A script to be parsed, a novel to be adapted and text on a reference image all have to be read, and nobody publishes anything about any of them.The only documented inputA recording to be dubbedHas to be recognised. One published list, of 72 entries.A script to be parsedHas to be understood. No vendor publishes which languages.A novel to be adaptedThe whole adaptation depends on it. Undocumented.Text on a reference imageHas to be read off a picture. Undocumented.Where a studio has to test rather than read
Fig. 1 A studio with source material in a less widely supported language cannot check anything before paying.
Where a source language matters, by kind of input. Recorded 2026-09-22.
InputWhy the source language mattersPublished by any vendor here
A recording to be dubbedIt has to be recognised firstOnce, as a 72-entry list
A script to be parsedIt has to be understoodNo
A novel to be adaptedThe whole adaptation depends on itNo
A reference image with text on itThe text has to be readNo

Inclusion rule. The kinds of input the products read here accept, with whether any vendor publishes which languages it handles. Order. Fixed order, the input most often documented first.

1The input question is invisible in most language claims

A claim about how many languages a tool supports is almost always about output. But a studio with a catalogue of scripts in one language needs to know the tool will read them, which is a different list.

One vendor here publishes an input list, for recordings. Nobody publishes one for text, so a studio with source material in a less widely supported language cannot check anything before paying.

2Input and output lists can differ in either direction

A dubbing tool usually hears more languages than it speaks. A generation tool might accept a script in one language and produce documents in another, which is a feature rather than a gap.

One vendor here states that its generated documents follow the interface setting, which implies input and output can differ deliberately. It is the only statement of that kind in the register.

3Why this register has no input column

Because one vendor publishing one list is not a column. A column needs enough vendors saying comparable things that an empty cell means something, and a column with a single entry would be a question dressed as a measurement.

The one published input list is recorded in the voice column beside the output list it belongs with, and the absence everywhere else is noted rather than tabulated.

A definition, not a measurement: no vendor figure appears here that is not also in the register. The register keeps three columns and explains the choice on the columns page.

Nearby terms: Speech recognition, Target language. The whole vocabulary is at terms; nothing on this page is a claim about a named product.