SBV to SRT Converter

SBV is the format YouTube’s own caption editor hands back. It is the simplest format in this section: two timestamps, then the lines, then a blank line. No header, no index numbers, no styling and nowhere to put any. Which is exactly why an SBV file turns up so often in front of a tool that has never heard of it — a caption download from YouTube is where most people first meet the extension, and the reason to convert SBV at all is that almost nothing else reads it.

YouTube captions are the source of nearly every one of these files, and they arrive in two states: written by a person, or generated automatically. Both are SBV, and the conversion is the same for either.

The comma is in a different place, and that is the whole problem. An SBV timing line reads 0:00:01.000,0:00:04.000: a full stop before the milliseconds, and a comma separating the start from the end. A SubRip timing line reads 00:00:01,000 --> 00:00:04,000: a comma before the milliseconds, and an arrow separating the two. Both formats have a comma between two numbers, so a tool that looks for the comma rather than for what sits either side of it reads one as the other.

When that happens, the file does not fail — it half works. Every pair of cues is read as one, each running from its own start to the next cue’s, and the result is a track with half the lines and every one of them on screen twice as long as it should be. What actually distinguishes the two formats is the arrow: SBV has none, and neither SubRip nor WebVTT can express a cue without one.

Index numbers are added, not preserved. SBV has none at all. SubRip requires a sequential number above every cue, so the output is numbered from one. WebVTT does not require them and does not get them.

The hours are single-digit and stay optional going in. SBV writes 0:00:01.000 rather than 00:00:01.000. SubRip requires the long form, so the output always writes it, whichever way the input was written.

Press Load a sample. It is three cues of SBV, and the status line reads read as SBV — decided from the content, not from the extension, so it also answers the question of whether the file is what its name claims.

Convert it to SubRip. The first cue becomes 00:00:01,000 --> 00:00:04,000 with a 1 above it: the hours padded to two digits, the comma moved from between the timestamps to inside each one, the arrow added, and the index that SBV never had written in.

The second cue’s two lines stay two lines. Where a caption breaks is a decision somebody made about where the sense breaks, and re-wrapping it silently would throw that decision away.

Now convert the same input to WebVTT instead. The WEBVTT header appears at the top, the commas inside the timestamps become full stops, and the index numbers are not written — WebVTT does not ask for them.

Automatic captions from YouTube usually need one more pass after this: turning them into a transcript collapses the repeated line that rolling captions leave behind, and tidying the timings separates the cues that were written to overlap on purpose.