AreaAudit

Which languages each generator says it supports, and where

Why an empty cell is a finding rather than a gap

A blank cell in a table like this means the vendor publishes nothing on the question. It is not a finding that the capability is missing, and in two of these three columns it is the most common state. As of 2026-09-12.

What a blank cell asserts, and the three things it does notA blank asserts that the vendor publishes nothing on the question in a place a reader can open. It does not assert that the vendor said no, that the capability is absent, or that nobody looked, and each of those would need a different kind of evidence.What could a blank cell mean?Asserted hereNothing is publishedEstablished by readingthe vendor's own pageson a stated day.Not assertedThe vendor said noWould need a publishedstatement to thateffect, and none exists.Not assertedThe capability isabsentWould need a test of theproduct, which a readercould not repeat.Blanks are exported with every other row
Fig. 1 Recording the absence is what keeps the finding checkable; inferring from it would replace something verifiable with something nobody could repeat.
Three readings of a blank, and which one is being asserted. Recorded 2026-09-12.
ReadingAssertedWhat would establish it
The vendor publishes nothing on itYesReading the vendor pages
The vendor has said noNoA published statement to that effect
The capability does not existNoTesting the product
Nobody lookedNoThe reading date says when it was checked

Inclusion rule. The readings a blank admits, with the one a register built on published evidence can assert. The others would need evidence of a different kind. Order. The asserted reading first.

1Most of these products can do more than they say

Almost all of them generate audio, and several plainly generate it in more than one language. None of the vendors with an empty voice cell states which languages, anywhere a buyer can check before paying.

So the blank records a gap in disclosure, which is a real and checkable thing, rather than a gap in capability, which would need a test. Filling it from a product description or a demo would replace a checkable absence with an unrepeatable one.

2Dropping the blanks would make the field look better documented

A table of only the filled cells would be much shorter and much more flattering. Half the vendors publish nothing about spoken output and all but one publish nothing about generated documents.

A reader scanning such a table would conclude that the questions have mostly been answered. The blanks are what show that they have mostly not been asked, which is the more useful finding for anybody about to buy.

3A blank has to be dated like anything else

An absence from six months ago is a statement about six months ago. Language lists grow quietly, so a blank is the cell most likely to have changed since it was recorded.

That is why an absence is written as a phrase with a date in it rather than as a dash. The claim being made is visible on the face of the cell, and it is a claim about a day rather than about a product.

4Blanks belong in the export as well

A data file that contained only the positive claims would be smaller and would misrepresent the same way the table would. A market with no claims is the finding a buyer in that market most needs.

So absence rows are exported with everything else, carrying the same source and date columns. Anybody rebuilding the comparison gets the blanks or gets nothing comparable.

An explainer about method and structure, not a count. No vendor number appears here that is not also in the register. The sourced material is on the register. Related: Two-sided lists, When the choice is made.