SRT, ASS, VTT — what's the difference between subtitle formats?

Every text subtitle is fundamentally a table: this line is shown from this second to that second. Formats differ in how much else the table can hold.

SRT: the simplest ​

12
00:01:04,520 --> 00:01:07,180
魔法使いといえば、たったひとつ。

Index, start and end time, text — that's all. No fonts, no positions; how it looks is entirely up to the player. Because it carries nothing extra it works anywhere and anyone can edit it — fixing a typo or shifting everything by two seconds is a job for Notepad.

Jimaku's Japanese subtitles and subtitles extracted from Netflix and other streaming services are mostly SRT.

ASS: the fansub standard ​

[V4+ Styles]
Style: Default,Noto Sans JP,60,&H00FFFFFF,...

[Events]
Dialogue: 0,0:01:04.52,0:01:07.18,Default,,0,0,0,,{\an8}魔法使いといえば、たったひとつ。

An ASS file first defines styles (font, size, colour, outline, shadow) and each line references one; the text can also contain override tags like {\an8} — place the line at the top of the screen, change colour, fade in and out, karaoke colouring character by character. Translating signs, text messages and OP/ED lyrics all rely on this, so nearly every fansub release is ASS. .ssa is its older version; any player that reads one reads both.

The price is that it demands more of the player: one ASS file may look different in different players, and a missing font falls back to a default.

VTT: SRT for the web ​

WEBVTT

00:01:04.520 --> 00:01:07.180
魔法使いといえば、たったひとつ。

Almost identical to SRT, designed for the browser <video> element (a dot instead of a comma in timings, a WEBVTT header). Web players and the subtitle tracks that come with Aniyomi online sources are almost always VTT.

What's the difference in Fushi? ​

None that matters — Fushi reads all of them, whether as external files or embedded mkv tracks.

The first thing Fushi does with a subtitle is separate styling from text: lookups, mining and what's written to Anki are plain text; tags like {\an8} never leak into a sentence. Part of the ASS styling is rendered (alignment \an, \pos, italic and bold, colour, size, line breaks); karaoke, animation and vector drawing are silently dropped — Fushi's subtitles are a layer of tappable characters and can't be painted into the picture the way mpv does, or the tapping would stop working.

So for Japanese, don't choose by format — choose by content: one Japanese subtitle whose lines match the dialogue is worth more than any effect. A fansub group's beautifully styled English ASS is only ever a reference; mining needs the Japanese one.

Aside: graphic subtitles ​

PGS (Blu-ray) and VobSub (DVD) aren't text but bitmaps and have no "format" to speak of — Fushi displays them and, once you pause, OCRs them so you can tap to look up, or converts the whole track to a text subtitle with OCR; see external, embedded and burned-in subtitles.