Convert SUB to VTT

The situation is a collection meeting a website. A lecture series, a local television archive, a film library, a set of recordings somebody has been carrying since the DivX era — the video gets transcoded and put online, and the subtitles that came with it are MicroDVD, which is what everybody used when the files were made.

A browser has no idea what to do with them. The track element accepts WebVTT and only WebVTT, and it fails silently: the video plays, the subtitle menu is empty, and nothing in the page or the console explains why. Between the two formats sit twelve years and a change in what a subtitle file is fundamentally measuring.

MicroDVD records the frame a line appears on, not the moment. Turning that into a timestamp is a division, and the divisor is the frame rate of the video the subtitles were made for — which the file rarely states and which the converter cannot infer. The control here accepts 1 to 120 in thousandths, because 23.976 and 29.97 are not whole numbers.

Get it right before publishing rather than after, because a browser gives you no diagnostic to work with. A player at least lets you nudge a subtitle offset while watching; a track element renders exactly what the file says. Check the converted track against the video locally, at the beginning and near the end, before it goes anywhere near a page.

There is a second reason to be careful here that does not apply to a subtitle file on a hard disk. A web page is often the version of the material that gets linked, quoted and indexed, and a caption track is text a search engine and a screen reader both read. Captions that are four minutes adrift by the end of a recording are not merely annoying to watch — they attach the wrong words to the wrong moment for every reader who never sees the video at all.

Nothing about the two formats lines up. WebVTT requires the word WEBVTT on its own first line; MicroDVD has no header of any kind. WebVTT separates two timestamps with an arrow on their own line; MicroDVD puts two frame numbers in curly braces at the front of the text line. WebVTT states times; MicroDVD states frames.

So the extension is the least of it. A renamed file is rejected at the first line, and even if it were not, there is nothing in it a WebVTT parser could interpret as a cue. This is the one pair on the site where the two formats share no syntax at all, which is also why the conversion has to be a real parse rather than a search and replace.

Each frame number is divided by the chosen rate and written as hours, minutes, seconds and three digits of milliseconds, with a full stop before the fraction and the hours zero-padded. A line beginning at frame 3210 starts at 00:02:08.400 if the rate is 25, and at 00:02:13.875 if the rate is 23.976.

Those five and a half seconds are the whole weight of the setting, at the two-minute mark of a film that may run for two hours. The timestamps rarely end in round numbers because the division rarely does, which is a reliable tell that a WebVTT track began life as a frame-based file. It has no effect on playback: a browser reads milliseconds and does not care where they came from.

Many MicroDVD files declare their frame rate using an old convention: a first entry spanning frame one to frame one whose text is the rate. It was never part of a specification and support for it was always partial, which is why this converter treats it as what it structurally is — an ordinary entry.

The result is that your WebVTT track opens with a caption reading something like 23.976, on screen for a fraction of a second. On a page that is a visible defect in the first moment of playback, which is the worst possible place for one. Open the .vtt in a text editor, delete the first cue, and the rest of the file is untouched.

MicroDVD has a small set of inline codes for italic, bold, colour, font and size, written in curly braces. They are removed along with the braces, leaving the words. Nothing is translated into the WebVTT equivalents, even though WebVTT has angle-bracket tags for italic and bold that would map cleanly.

That is worth knowing rather than worth mourning. Player support for the MicroDVD codes was inconsistent for the format’s entire life, so a file relying on them looked different everywhere, and most archive files use them for one thing: marking a whole line as italic where a character is speaking off screen. If that distinction matters on your page, restoring it is a pass with a search and replace, and the WebVTT form is the one every browser honours.

Put the .vtt in a track element inside the video element, with kind set, srclang set to the language code, and a label that becomes the entry in the subtitle menu. Add default if the subtitles should be on when playback starts, because otherwise a viewer has to find the menu and most will not.

Language is the field archive material gets wrong most often, because a file inherited from a collection frequently has no record of what language it is in beyond the folder it was found in. Declaring it correctly is not decoration: it is what tells a browser which track to select for a viewer’s preferences, and what tells a search engine that a page carries a transcript in a particular language.

MicroDVD belongs to an era when subtitle files were single-byte and regional, and archive files are commonly saved in a Central European, Cyrillic or Western European encoding rather than UTF-8. The converter writes UTF-8 output, which is what a browser expects, but it can only encode the characters it read.

So the check happens before conversion, not after. Open the SUB in a text editor and look at the accented characters. If they are already question marks or stray symbols, the file is being read with the wrong encoding, and re-saving it as UTF-8 from an editor that lets you choose the source encoding is the step that recovers them. A file converted while misread produces a WebVTT track that is wrong in the same way, permanently.

Two entirely different things share this extension, and an archive is exactly where you meet the second one. MicroDVD SUB is plain text. VobSub SUB is bitmap subtitles ripped from a DVD, stored as pictures, always with an .idx index file beside it. Open the file in a text editor: readable lines in curly braces mean MicroDVD, binary noise with an .idx alongside means VobSub.

For a web publisher the distinction is a budget item rather than a curiosity. There is no path from images of words to a WebVTT track that does not go through optical character recognition, which means a separate tool, a review pass on every line, and accuracy that depends on the quality of the original transfer. Discovering it at the point of putting the video live is worse than discovering it while planning, and the check takes ten seconds per file.

The kind attribute distinguishes them and the distinction matters more on the web than it did in a media player. Subtitles carry the dialogue and assume a viewer who can hear. Captions assume a viewer who cannot, and include what is audible but not spoken: who is speaking off screen, when music starts, what the sound was.

A MicroDVD file from an archive is essentially always subtitles, often a translation. Marking the track as captions to satisfy an accessibility checklist puts the right word in the markup and does not meet the requirement, which for public sector bodies is a legal one rather than a preference. If captions are needed, the non-speech information has to be written by somebody watching the video, and no conversion produces it.