Convert SUB to SBV

Files with the MicroDVD extension are old — the format belongs to the DivX-player era and almost nothing has produced one for fifteen years. They turn up alongside archive video: a ripped collection being digitised properly, a lecture series recovered from a shelf of discs, a film society putting its catalogue online. The video gets re-encoded and uploaded; the subtitles have to come with it.

YouTube is the destination often enough for this pair to be worth writing, and it is a friendlier one than a disc player. A player takes the file and shows whatever it makes of it. YouTube stores the cues, renders them with its own styling, and lets you edit the timings afterwards in the browser — which changes how much precision has to be right on the first attempt.

MicroDVD does not say when a line appears. It says which frame it appears on: `{3210}{3455}` and then the text. A frame number is only a moment in time if you know how many frames there are in a second, so the conversion cannot proceed without one, and the field on this page is where you supply it.

The default is 25, which is the European broadcast rate and the likeliest value for material of this age. The other strong candidate is 23.976, the film rate, used by anything mastered from a cinema source. The field accepts thousandths of a frame precisely so that 23.976 and 29.970 can be entered exactly rather than rounded. Get it right and the conversion is exact; get it wrong and every line is displaced by an amount that grows as the film goes on.

Open it in a text editor first. There is a long-standing convention that the first entry of a MicroDVD file has both frame numbers set to one and its text is the frame rate — `{1}{1}23.976` — which players read as a declaration rather than displaying. Many files have it. If yours does, that is the answer and there is nothing to work out.

If it does not, the arithmetic is short. Take the last cue’s starting frame number, divide by a candidate rate, and see whether the result lands near the end of the film. A file whose final frame number is 129,600 divided by 25 gives 5,184 seconds, which is eighty-six minutes; the same number divided by 23.976 gives ninety minutes. Compare each against the running length you know and the right one is usually obvious.

This is the diagnostic worth learning because it separates two problems with completely different fixes. If the subtitles are wrong by a constant amount from beginning to end — two seconds early at the start and two seconds early at the end — the timing base is right and the file needs shifting, which YouTube Studio can do after upload.

If the error grows steadily, barely visible at the start and minutes out by the end, the timing is being scaled wrongly and that means the frame rate. The classic pair is 25 against 23.976: the ratio is 1.0427, so the error is 4.3 per cent of however far into the film you are — about twenty-six seconds after ten minutes and nearly four minutes by the end of a ninety-minute feature. Imperceptible in the first thirty seconds, unwatchable in the third act. Convert again with the other rate rather than trying to nudge it.

If the file did carry that `{1}{1}23.976` declaration, the conversion has no way to know it was meant as metadata — it is a well-formed entry at frame one — so it becomes the first cue in the SBV, a fraction of a second long, whose text is a number.

Delete it. The SBV file is plain text with one block per cue, so removing the first block and the blank line after it takes a moment in any editor. Doing it before upload is better than doing it in YouTube Studio afterwards, because a stray caption at 0:00:00 is exactly the sort of thing nobody notices until a viewer mentions it.

A MicroDVD entry is one line: two frame numbers in curly braces and then the text, with the pipe character standing in for a line break. An SBV entry is a timing line — two `H:MM:SS.mmm` values separated by a comma, with the hour unpadded — followed by the text on its own lines, and a blank line between entries.

So the conversion turns one dense line into a three-line block, expands the pipes into real line breaks, and replaces frame counts with millisecond timestamps. The result is several times the size of the source and far easier to read and correct by hand, which matters given that an archive subtitle file usually needs some correcting. Cue order follows the file rather than the clock, so an out-of-order source stays out of order.

MicroDVD has a small set of inline codes written in curly braces at the start of a line: italic, bold, underline, colour, font and size. More than SubRip has, and applied inconsistently across the players of the era. All of them are removed, and the SBV carries plain text.

Nothing is lost in practice, because YouTube renders captions with its own typography and offers the viewer control over the font, size, colour and background. A styling instruction in an uploaded file would not be honoured even if the format could express it. What does survive is the line structure — a pipe becomes a real break, and YouTube shows the two lines as two lines.

YouTube accepts several caption formats, SubRip among them, and for a single upload SRT is the more portable choice by a distance. The honest answer to "should this be an SBV" is usually no.

The case where it is yes is a workflow rather than an upload. YouTube Studio exports SBV, so a channel whose caption archive, review process or correction script was built around what Studio produces has a good reason to keep everything in one format — round-tripping a file out of the editor, through a correction pass, and back in without a format change removes a class of mistake. If that describes you, this is the conversion. If you are adding captions to one video, convert to SRT instead and keep both.

This destination forgives more than a hardware player does. Once the file is uploaded, Studio’s caption editor shows the cues against the waveform and lets you drag or type new times, and there is a control to shift the whole track. So a residual error of a second or two — from a slightly different encode, or a source that had a few frames of leader — is a two-minute correction rather than a re-conversion.

What it will not fix cheaply is a rate error, because that means every cue is displaced by a different amount and there are hundreds of them. Correct the rate here and use Studio for the offset, which is the division of labour that makes this pair worth doing at all. Preview against the actual upload before publishing: the copy you uploaded may have been re-encoded or trimmed relative to the one the subtitles came from.

The SBV is a derived file and the MicroDVD original is the record of where the text came from. Keep it — if the video is ever re-encoded from a different source at a different rate, converting again with a new number is a ten-second job, whereas rescaling a timestamped file is not.

The conversion itself is plain JavaScript in this browser tab: no upload, no engine download, no account and no daily allowance, with the network tab as the check. Archive subtitle files are frequently somebody’s unpublished transcription work, and a file that never leaves the machine is one fewer thing to explain.