Convert SRT to ASS

A skeleton with correct bones. The words and the timings are the only things a SubRip file contains, and they are the only things that can be carried over, so the result is every one of your cues rendered as an ASS dialogue line under a single default style. Nothing has been invented, because there was nothing to invent from.

That is more useful than it sounds, and it is worth being clear about why. Timing is the slow, tedious half of subtitle work; styling is the fast, enjoyable half. If the timings are already right, this conversion moves the expensive part of the job across intact and leaves you at the point where the format becomes worth using.

One style, named Default, and these are its actual values: font Arial at size 48, primary colour white, a black outline two pixels thick, no shadow, and alignment 2 — the bottom-centre cell of the three-by-three grid the format positions by. Left, right and vertical margins are all 10. The script header declares ScriptType v4.00+, WrapStyle 0 and ScaledBorderAndShadow yes.

Knowing those numbers tells you what to change first. Arial at 48 is a deliberately unremarkable choice that renders everywhere; it is not a recommendation. If the track is going into a 1080p video and the text looks small, the size is the field to touch, and because ScaledBorderAndShadow is on, the outline scales with it rather than staying two pixels of the original resolution.

SubRip records milliseconds; ASS records hundredths of a second. That is a real difference in resolution, and the conversion resolves it by truncating rather than rounding — 00:00:01,238 in the SRT becomes 0:00:01.23 in the ASS, nine milliseconds early at worst.

Nine milliseconds is a quarter of a frame at 24 frames per second and nobody has ever noticed one. It matters in exactly one situation: karaoke, where syllable timing is set against the audio and the accumulated difference between adjacent syllables can be heard. If you are timing karaoke, do it in the ASS after conversion, not by converting a file that was timed elsewhere.

A subtitle split across two lines in the SRT arrives as one dialogue line containing a backslash and a capital N where the break was. That is the format’s hard line break, and it is how ASS distinguishes a break the author asked for from one the renderer chose.

The alternative in ASS is a soft break, a lower-case n, which the renderer may ignore when WrapStyle allows it to reflow. The conversion always writes the hard form because an SRT break was a decision somebody made, and silently letting a renderer undo it would be the wrong default. If you would rather the text reflowed to fit, replace them.

Alignment 2 puts subtitles at bottom centre, which is where subtitles belong and where the whole point of ASS begins. The format positions by a numeric grid, so 8 is top centre, 7 is top left, and an inline override tag can place a single line anywhere in the frame without touching the style.

That is the mechanism behind translated on-screen signs, which is what most typesetting time is spent on. A sign on a shop front, a text message, a caption in the source language — each becomes a dialogue line positioned over the thing it translates, in a style chosen to match it. None of that can come out of an SRT; all of it becomes possible the moment the file is ASS.

This is the one result that surprises people, so it is worth stating plainly. Cue text is copied verbatim. An angle-bracket italic tag that some SubRip players honour by convention is not converted into the ASS equivalent, which is a curly-brace override; it stays exactly as typed, and an ASS renderer will draw the angle brackets on screen as characters.

The fix is a search and replace over the finished file, or setting italic on the Default style if the whole track is italic. Either takes a minute. Doing it is preferable to a converter guessing, because those angle-bracket tags are a player convention rather than part of SubRip, and a file that used them for something else would be silently mangled by a translation step.

The style says Arial. It does not carry Arial. An ASS file references fonts by name and relies on the playback machine having them, which is the format’s defining weakness and the reason a carefully typeset track can arrive somewhere looking broken through no fault of the file.

A substituted font has different metrics, so a line that fitted now wraps, and a sign positioned against the picture now overlaps it. The three known answers are to stay with fonts that exist everywhere, to ship the fonts and ask the viewer to install them, or to attach them inside a Matroska container, which is what release groups do. Arial is the default here precisely because it is the least likely to be missing.

The decision is easier than it looks once you know where the file is going. A track for your own machine can use anything installed on it. A track being sent to one other person can travel with its fonts and an instruction. A track being published, or destined to be burned into the picture, should either restrict itself to what is universal or stop being a subtitle file at all — because the only way to guarantee a typeface renders on a stranger’s machine is for it to have been rendered on yours first.

Aegisub is the program this format grew up with and it remains where serious subtitle work happens: styles managed in a table, timing set against an audio waveform, and typesetting tools for placing and transforming individual lines. The file this converter produces opens in it directly and cleanly.

A text editor also works for small corrections, since the events section is comma-separated plain text. Leave the styles section alone unless you know what a field does — several of them are colours written in reverse byte order behind an ampersand-H prefix, which is not a thing worth discovering by experiment on a file you care about.

The conversion writes one style and applies it to every line, which is the only honest thing it can do from a SubRip source. The first edit that repays the move is usually a second style. Duplicate Default in the styles table, rename it, change the primary colour, and reassign the lines that belong to the other speaker — the style name is a field on each dialogue line, so reassigning is a column edit rather than a rewrite.

Colour-coding speakers is the oldest trick in subtitling and it works because it costs the viewer nothing to learn. It is also the thing SubRip could never do: a plain file has to write the speaker’s name into the text or use a dash convention, both of which spend characters on the screen that the dialogue needed. Two styles buy back a line of reading space in every exchange, and in a conversation-heavy piece that is the difference between captions that keep up and captions that do not.

If nothing about the track needs to be positioned, coloured or styled, converting is work with no return. SubRip is read by every player, every television and every upload form; ASS is read by desktop players and almost nothing else, and a platform handed one will usually ignore it.

The rule that holds up is destination-driven. Typeset in ASS when the video will be watched in a player that renders it, or when you intend to burn the subtitles into the picture and the appearance is yours to decide. Stay in SRT when the file is going somewhere you do not control. Both can be true of the same project, in which case the ASS is the master and the SRT is generated from it.