Convert ASS to SRT

Two things survive: the text of every dialogue line and the moment it starts and stops. Everything else is a property of the format you are leaving. The style table, the font names, the primary and outline colours, the alignment, the margins, the per-line overrides, the fades, the moves and the karaoke timing all have nowhere to go, because SubRip contains no syntax that could hold any of them.

The registry declares this pair approximate rather than lossless, and the label is accurate. It is not the only one: reading a WebVTT file drops its cue settings, and anything touching MicroDVD depends on a frame rate that may be wrong. Converting between SubRip and the YouTube format moves the same information into different punctuation. This conversion throws away the reason somebody chose the format, and it should be done with a copy rather than in place.

Everything between curly braces is removed. A positioning tag, a colour change mid-sentence, a fade, a rotation, a blur — all of it disappears cleanly rather than being drawn on screen as characters, which is what happens when a file is renamed instead of converted.

Two escapes are translated rather than deleted. The hard line break, a backslash and a capital N, becomes a real newline so a two-line subtitle stays two lines. The hard space becomes an ordinary space. Those are the only pieces of ASS syntax that mean something in SubRip terms, and they are the only ones carried over.

Typesetters draw shapes in ASS: a box behind a translated sign, a mask over on-screen text, an arrow. The drawing is switched on by a tag in braces, and the coordinates that follow it sit in the dialogue text as ordinary characters. Only the tag is in braces, so only the tag is stripped.

What lands in the SRT is a caption reading as a run of letters and numbers — the drawing commands themselves. It is unmistakable when you see it and impossible for a converter to distinguish from legitimate text without interpreting the format it just left. Scan the output for cues that are all numbers and delete them; on a heavily typeset release there may be a few dozen.

They cluster where you would expect. A drawing is usually a mask or a backing shape behind a translated sign, so the coordinate cue and the sign it belonged to sit next to each other in the file and cover the same seconds. That is a useful pattern when you are cleaning up: find one, and the caption immediately around it is almost always the sign, which is itself a candidate for removal on a plain track.

Karaoke in ASS is a tag before each syllable saying how long that syllable lasts, which is what lets a player highlight the line as it is sung. The tags are in braces, so they are stripped, and the syllables between them join back into an ordinary line of text.

The result is correct and disappointing: the whole lyric appears at the start of the line and disappears at the end, like any other subtitle. Nothing is mistimed and nothing is garbled — the effect simply does not exist any more. If the karaoke was the point of the file, this conversion is not the answer and burning the subtitles into the video is.

An ASS file routinely has two or more lines on screen at once, on separate layers: the dialogue at the bottom and a translated sign positioned over the thing it translates. That is normal and it is one of the things the format was built for.

SubRip has no layers and no positions, so both lines become cues that cover the same seconds. Overlapping cues are not illegal in SubRip and many players will stack them, but behaviour varies and some show only one. If the output looks like it is dropping lines, this is usually why, and the fix is editorial: decide which of the two the viewer needs and remove the other.

For a track being delivered to a platform, the dialogue is nearly always the one to keep. A sign translation that appears under the picture, out of position and with no visual link to the thing it translates, reads as an unattributed fragment in the middle of a conversation — it costs the viewer more attention than it returns. Where a sign genuinely carries the plot, folding its text into the adjacent dialogue cue in brackets is the convention broadcast subtitling settled on decades ago.

An events section contains Dialogue lines and often Comment lines, which are structurally identical but disabled — alternate translations, notes between team members, a sign that was typeset and then cut. Only Dialogue lines are read, so none of that reaches the SRT.

That is the right default by a wide margin, and it is worth knowing about because it works in both directions. If a line you expected is missing from the output, open the ASS and check whether it says Comment. Re-enabling it in Aegisub and converting again is a two-second fix, and considerably faster than concluding the parser lost it.

It is also the reason a converted file is sometimes shorter than the source suggests. A translation that went through review can carry as many commented alternates as live lines, and a script that looks like twelve hundred entries in an editor becomes seven hundred cues in the SRT. Count the Dialogue lines rather than the total if you want to know in advance what the output will contain.

The events section declares its own column order in a Format line, and real files genuinely differ — the order is a convention rather than a fixed layout. Reading the start, end and text by fixed position works until it meets a file written by a tool that ordered them differently, and then it produces silent nonsense.

The columns here are located by name from that Format line, so a file with an unusual layout parses correctly rather than yielding cues timed from the wrong field. The text column is read as everything from its position to the end of the line, which matters because dialogue containing commas would otherwise be truncated at the first one.

Cues are numbered from one in the order they appear in the events section. Aegisub keeps that section sorted by time most of the time, so most files come out chronological and there is nothing to think about.

Heavily typeset scripts are the exception. A sign inserted while working on a scene often lands next to the line it belongs with rather than at its place in the timeline, and the SRT then contains cues whose timestamps run backwards relative to their numbers. Most players sort by time and do not care; strict tools and some upload validators reject the file. Sorting the events section by start time in Aegisub before converting avoids it entirely.

ASS timestamps carry two digits after the decimal point; SubRip carries three. The conversion pads rather than guesses, so 0:00:01.23 becomes 00:00:01,230. No precision is invented and none is lost — every timestamp in the output simply ends in a zero.

That is a small tell that a file came from ASS, and occasionally a useful one when you are trying to work out where a subtitle track has been. It has no effect on playback: ten milliseconds is a quarter of a frame at 24 frames per second, and the resolution the ASS was authored at is the resolution the timings genuinely have.

Nothing that was stripped can be recovered from the output. The SRT does not know which lines were signs, what colour a speaker was, where a caption sat or how a syllable was timed, and no amount of later work will reconstruct it. The ASS is the only copy of that, which makes deleting it after a conversion an unusually expensive kind of tidying.

The working arrangement is the one release groups already use: the ASS is the project, and every plain file is an export. When a translation changes, change it in the ASS and convert again. Two files edited in parallel diverge quickly, and the plain one always wins by accident, because it is the one that opens fastest.