Set subtitle size as a share of the video's height, not in points. suubbs' default is 5.2% — 56 px on a 1080p video, 100 px on a 1080×1920 vertical one — and most of its looks sit between 4.8% and 6.6%.
A font size in pixels or points means something different on every video. A fraction of the frame means the same thing on all of them.
"48 pt" is large on a 720p clip and small on a 4K one, and it is a different size again on a vertical video shot at 1080×1920. That is why people asking "what font size for subtitles" get contradictory answers: the number only means something next to a resolution. The fix is to think of size the way the picture does — as a share of the frame's height — and convert to pixels only at the last step, for whichever file or tool needs pixels.
That is how suubbs stores it. The Size control in its style panel runs from 2.4% to 10% of the video's height, in steps of 0.2, and the same caption therefore looks the same on any resolution you open. The built-in looks give a sense of the useful range:
| Look | Size | Face and treatment |
|---|---|---|
| Film | 4.8% | Georgia italic, regular weight, soft shadow |
| Classic | 5.0% | Inter semibold, white, soft shadow |
| Boxed (the default) | 5.2% | Inter semibold, white on a near-black panel |
| Highlight | 5.2% | Inter bold, dark text on a lime panel |
| Pop | 6.6% | Inter Tight black, upper case, dark outline |
Read the pattern rather than the exact figures. Quiet, mixed-case looks meant to be read over a long video sit near 5%. The loud, all-capitals social look goes larger, because it shows a few words at a time and is meant to be the loudest thing on screen. As a rough guide rather than a rule: much below 4% the type starts to ask for a large screen, and much above 7% it reads as a headline rather than a subtitle, with two lines of it covering a good part of the picture.
Percentage of height multiplied by the height, rounded — exactly the arithmetic suubbs does when a format needs a pixel figure.
| Size | 720p (720 tall) | 1080p (1080 tall) | Vertical 1080×1920 | 4K (2160 tall) |
|---|---|---|---|---|
| 2.4% (smallest) | 17 px | 26 px | 46 px | 52 px |
| 4.8% | 35 px | 52 px | 92 px | 104 px |
| 5.2% (default) | 37 px | 56 px | 100 px | 112 px |
| 6.6% | 48 px | 71 px | 127 px | 143 px |
| 10% (largest) | 72 px | 108 px | 192 px | 216 px |
Note the vertical column. A 9:16 video is 1,920 pixels tall, so the same 5.2% is almost twice as many pixels as on a 1080p widescreen video — but the frame is only 1,080 pixels wide. Size measured against the height keeps the type in proportion to the picture; it does not give you more room across. That is the real constraint on vertical captions, and the next section is about it.
Font size, line length and the width of the caption box are one decision, not three.
A widescreen subtitle line traditionally holds up to 42 characters over at most two lines, and that is suubbs' default layout. On a 9:16 frame at a size around 5% of the height, only about 20 characters fit across, which is why its Vertical / social layout breaks lines at 22 characters and allows three. Push the size up to 6.6% and even 22 characters may not fit, so the line wraps on its own.
Rather than fight that, choose which one gives. In suubbs the caption box's Max width (40% to 100% of the frame) and the type size decide where a caption wraps in the preview and in an ASS burn-in; the character limit itself is set by the line layout, Widescreen or Vertical / social. If a larger size leaves too many lines, show fewer words at once — a low Words on screen setting gives the few-words-at-a-time look of word-by-word captions, which is why that style can afford big type. The reading-speed side of the trade, and where the 42 and 22 figures come from, is on subtitle characters per line and reading speed. On vertical video, the width you can use is also limited by the apps' own buttons: see caption safe zones for TikTok, Reels and Shorts.
The size you chose only reaches the viewer if the file can carry it. Each format handles it differently.
Take the Pop look — Inter Tight at 6.6% — on a 1080×1920 video, and export it from suubbs four ways.
An ASS file declares the resolution it was written against (PlayResX and
PlayResY), and suubbs writes the video's own dimensions there, so every size and
margin in the file is literal. The style line carries the face and a font size of 127 — 6.6% of
1,920. When ffmpeg burns the file in through libass, the type comes out the size you previewed.
Outline and shadow thickness are scaled to the type size as well, so a heavier font does not end
up with a hairline outline. This is the file to use for burned-in captions; see
how to burn subtitles into a video.
SRT has no style block, but many players honour a small set of inline tags. suubbs writes the size as one, converted to pixels against the video's height:
1
00:00:01,000 --> 00:00:02,400
<font face="Inter Tight" size="127">MADE ON MY MAC</font>
Only what differs from suubbs' own default is written, so an unstyled document still exports
as the bare file every tool expects. Two caveats: a player that does not parse these tags shows
them as literal text, and how players interpret size varies, so test the one your
audience uses. More on what SRT can and cannot hold is in the
SRT file generator guide.
For a video editor rather than a player. Editors such as DaVinci Resolve and Premiere style a subtitle track as a whole, and a line that arrives carrying its own markup is treated as hand-styled and stops following the track. This export writes bare text, so you set the size once in the editor for every line. See subtitles in DaVinci Resolve.
WebVTT styles captions with CSS, but its ::cue rule resolves a percentage size
against the player's own default caption size, not the video. suubbs therefore writes size as a
ratio to a browser's usual default of about 5% of the video height — the Pop look becomes
font-size: 132%. A viewer who has changed caption size in their system settings will
shift that, which is arguably their right. Converting between the two text formats is covered in
SRT to VTT.
Size is only half of legibility. The other half is a face that holds up small and over moving footage.
suubbs offers five faces: Inter, Inter Tight, IBM Plex Mono, Georgia and Arial Black, in regular, semibold or black weights. The default is Inter at semibold — a plain sans with open letter shapes, set heavy enough that it does not thin out against a bright background. Georgia is there for the italic film look; Inter Tight and Arial Black for loud social captions where width is scarce and weight is the point.
One practical trap applies to burned-in ASS files. libass looks the font up by name on the machine doing the burn-in, and if it cannot find that family it quietly substitutes another, so the captions come out in a face you never chose. suubbs names its two bundled web faces exactly as their font files declare them and uses stock system names for the rest, but the machine running ffmpeg still needs the font available. If a burn-in looks wrong, check this before anything else.
Size and face also interact with the treatment behind the text. A white caption with no backing needs more size and weight to survive a bright shot than the same caption on a dark panel, which is one reason suubbs' default is the boxed look.
Around 52–56 px for ordinary subtitles, which is 4.8–5.2% of a 1,080-pixel height. suubbs' default of 5.2% works out to 56 px. Louder social-style captions run larger, around 71 px at 6.6%.
On a 1080×1920 frame, the same 5.2% is 100 px, and an all-capitals social look at 6.6% is about 127 px. The limit is width rather than height: keep lines short and inside the area the apps' buttons do not cover.
Many players pick their own caption size for SRT, and support for the size tag varies between them. If the exact size matters, burn the captions in from an ASS file, or set the size in your editor's subtitle track.
For a separate caption file, often yes: a browser applies the viewer's own caption-size setting to WebVTT, and many players offer caption settings of their own. For burned-in captions, no — the text is part of the picture.
No — it stays the same share of the frame. suubbs stores size as a share of the video's height and converts it to pixels at export, so the caption occupies the same share of the frame at any resolution.
A semibold or bold weight usually reads better over footage than a regular one. suubbs' default Inter is semibold; its film look is the exception, using regular-weight Georgia italic with a soft shadow.