About SubtitleMatrix

SubtitleMatrix is a set of tools for people who work with subtitles and the media they come from: converting between formats, fixing timing, cleaning up text, checking a file for defects, checking it against accessibility requirements such as reading speed and line length, and pulling subtitles out of video, bitmap files or audio or turning them back into speech.

It is a tool collection, not a service and not a publication. There is no account, no dashboard, no queue and no upload button. You open a page, hand it a file, and get a file back.

Most subtitle work is a text transformation, and a text transformation does not need a server, a queue, a database or a GPU. The media pages do heavier work — extracting tracks from a video, reading bitmap subtitles with OCR, transcribing speech, synthesising narration — but that work runs in WebAssembly engines downloaded into the tab, not on a machine we control. Almost every other subtitle site runs its work on a server anyway, which is why those sites have to ask you to upload your file first — the file has to be somewhere their code can reach it.

This site does not have that problem, because there is nothing on the other end. The parsing, the transformation and the file you download all happen inside the page. That is not a privacy setting that could be turned off: there is no endpoint to send your subtitles to, so there is no code path that could.

Every pair of formats has a page of its own, and each of those pages says which features survive the conversion and which do not. That list is computed from a capability matrix — a table of what each format can represent — rather than written by hand per page, so it cannot drift away from what the converters actually do.

The result is that a conversion page is not a generic tool with two dropdowns. It is a page about one specific pair, with its own explanation, its own worked examples and its own answers to the questions that pair raises.

There is no editorial byline to point at, and that is deliberate: the capability matrix, the loss reports and the compliance thresholds are all stated on the pages themselves, in terms you can check against the source formats and the published standards. If a claim here is wrong, the page shows you exactly what it claimed, which is what makes it correctable.

A loss report states what the capability matrix says about two formats. It is a lower bound, not a guarantee of completeness: a particular file can use features that the format allows and the matrix does not model, and those will not appear in the list.

A compliance check reports what the built-in rules can detect in the file you gave it. It is not legal advice and it is not a certification. Before a file is relied on to meet an obligation, have someone qualified review it. The media pages are a related case: OCR and speech recognition produce a best effort, so the output has to be read and corrected rather than trusted as transcribed. The terms page states both of these more formally.