Sentioscope

Speech and voice controls, as each vendor documents them

Why a number of languages settles nothing

Commissioning conversations do not ask how many languages a tool speaks. They ask whether it speaks one particular language. A total cannot answer that and a list can, whichever of the two numbers is larger. As of 2026-09-12.

Which market are you in, and which question answers itA commissioning conversation asks about one language. A total cannot answer that; a list can, whichever of the two numbers is larger. Size matters in one direction only.A list of sixteenA total of ninety-fiveIs my next market in itAnswered in a minuteCannot be answeredCan software check itWith codes, yesNoWhat the number describesVoices that existSounds alignment copes withWhen it goes staleVisibly, one name at a timeSilently, as the figure is reusedInterface languages and dialogue languages are two lists
Fig. 1 With a very large total, coverage can be assumed and verified cheaply; with a total of five the same assumption is a coin toss.
What each form of language answer settles, and what it leaves open. Recorded 2026-09-12.
Form of answerSettlesLeaves open
A claim of breadthThat the capability is intendedEverything else
A code per voiceWhich voices reach which market, automaticallyHow the delivery sounds
A list of namesWhether the next market is in scopeWhether it is usable there
A totalThat the capability is broadWhether it covers you

Inclusion rule. Forms in which a language capability can be published, from the most to the least specific. Order. Alphabetical by form of answer.

1A small list beats a large number

A named set of sixteen is more useful than an unnamed total of ninety-five, because one can be checked against a distribution plan in a minute and the other cannot be checked at all. Ranking the larger number higher would be ranking the wrong thing.

Size matters in one direction only. With a very large total a production can reasonably assume coverage and verify cheaply; with a total of five, the same assumption is a coin toss made before a season is committed.

2Interface languages and dialogue languages are two lists

Product pages routinely carry a language list belonging to the menus, the documentation or a prompt parser, and that list is longer than the one the actors speak. Reading the two as one is the commonest way this question gets answered wrongly.

The check is simple and worth making explicit in an evaluation: ask which list is being quoted. A vendor that cannot say has not distinguished them internally either.

3Named is still not the same as shippable

A named language moves a market from impossible to testable. Whether the delivery convinces somebody who speaks it at home is decided by a listener, and register, accent and idiom are documented by nobody.

So a list is where the work starts. One short exchange per market, heard by somebody who lives there, and kept as a finding beside the shot list, is the step a list makes possible and a total does not.

4Where the published detail is logged

What is described above is how the techniques work in general. Which models document which of them, in whose words, is kept in the speech controls table with a date on every field.

A mechanism note rather than a documented field. Nothing here is attributed to a model, and nothing here is a claim about one. The sourced material is on the speech controls table. Related: Two speakers, Keeping the clip.