This is a copy. Use Snipmill at snipmill.com.

Skip to content

Captions from what is said in the video, edited by you and burned into the picture.

Settings

Subtitles

The subtitles will appear here.

Press Create subtitles to start.

Speech

Look

Style White letters, dark outline

Create or upload subtitles first

Position Near the bottom edge

Create or upload subtitles first

Size The usual size

Create or upload subtitles first

Highlight the word being said The whole line in one colour

Create or upload subtitles first

Colour

Create or upload subtitles first

Runs on Snipmill's server: only the video's sound is sent, and deleted after. How we keep it private

Good to know

What adding subtitles does to your video

Burning in means drawing each line into the frames it belongs to, so the video is encoded again: the lines become part of the picture.

Picture
What happens: Every frame is encoded again with the line showing at that moment drawn into it, usually at about the original's bitrate (a little more, for the letters' edges)
Frame size, length, frame rate
What happens: Kept as filmed. Only the Captions for Reels page can cut the frame to 9:16, 1:1 or 4:5
Long lines
What happens: Split into captions of about two lines each, cut after a comma or full stop where it can, each shown for its share of the time. A caption that still needs more than three lines gets smaller letters instead of leaving the frame
Sound
What happens: In the MP4 as AAC: AAC sound is usually copied as it is, other sound is encoded to AAC
Format
What happens: MP4 with H.264 video, named after your file with "-subtitled"
Your original file
What happens: Unchanged

Which style and size to pick

Letters scale with the frame, so a 4K video and a phone clip look the same. At Medium size, on a 1080p video (the font is Inter ExtraBold, served by this site; Bangla, Hindi and other Devanagari, Arabic, Hebrew, Chinese, Japanese and Korean lines use Noto Sans ExtraBold, also from this site, and other scripts your device's font):

  • Small makes the letters 81% of Medium, Large 125%. Bigger letters fit fewer words, so captions change more often.
  • Colour: White, Yellow, Green, Cyan or Pink replace the style's own colour (yellow for Bold, white for the others).
  • Highlight the word being said: Colour turns it yellow (green when the letters are yellow); Grow also draws it 15% bigger, the rest of the line making room, and only colours a word that would push the line past the caption's width. Off keeps the line in one colour.
  • The word's time is the line's time shared out by word length (the speech model gives times per line, not per word), so a pause inside a line puts the highlight a little ahead. Right-to-left lines light up from the right; a line mixing right-to-left and left-to-right words lights up whole.
Clean
Look: White letters with a dark outline
Font size, wide 1920×1080: 59 px
Font size, vertical 1080×1920: 62 px
Characters per caption, wide / vertical: about 84 / 44
Box
Look: White letters on a dark box, like TV subtitles
Font size, wide 1920×1080: 59 px
Font size, vertical 1080×1920: 62 px
Characters per caption, wide / vertical: about 84 / 44
Bold
Look: Big yellow capitals with a thick outline
Font size, wide 1920×1080: 81 px
Font size, vertical 1080×1920: 85 px
Characters per caption, wide / vertical: about 49 / 24
Minimal
Look: Smaller white letters with a soft shadow
Font size, wide 1920×1080: 59 px
Font size, vertical 1080×1920: 62 px
Characters per caption, wide / vertical: about 84 / 44

Where the captions sit

Each position is a share of the frame's height, so it lands in the same place on a phone clip and a 4K video:

Bottom
Wide video (16:9): Last line ends 8% above the bottom edge
Vertical video (9:16): Last line ends 16% above the bottom edge
Middle
Wide video (16:9): Last line at 58% down the frame
Vertical video (9:16): Last line at 62% down the frame
Top
Wide video (16:9): First line starts 7% below the top edge
Vertical video (9:16): First line starts 10% below the top edge

Burned in, or a subtitle file

Burned-in captions are part of the picture: they show wherever the video plays, also with the sound off, and nobody can turn them off. A subtitle file (SRT or VTT) sits next to the video, and a player shows it only when the viewer turns subtitles on.

For a file, choose the same video in Audio to Text and save SRT or VTT: both hold every line with its start and end time, in the times of your video. SRT numbers each line and writes times as 00:01:02,450; VTT starts with WEBVTT and writes 00:01:02.450.

YouTube takes both formats: in YouTube Studio, open Subtitles, choose the video, press Add language, then Add under Subtitles, Upload file and "With timing".

How the speech becomes lines

A speech model on Snipmill's server, the same one Audio to Text uses, writes the lines. The page reads the video's sound, packs it small, and sends it; the picture isn't sent. The lines come back with their times, and the server deletes the sound. It works on a computer, phone or tablet. With your own subtitle file, the sound isn't sent.

It gives each piece of speech its start and end time. A part it hears as no speech, such as silence, noise or music, gets no lines, and looping text (a phrase of 4 or more words said 3 times or more, or text that keeps repeating itself) is dropped. How well it hears depends on the language, so read the lines before you burn them in.

The bar warns when the language heard is unsure, isn't the one you chose, or is mixed; when the file sounds like music; or when the text comes out in another language's letters. Where it heard another language than the one you chose, "Use <language>" runs it again in that one.

When the browser can't encode H.264

The video is usually made with your browser's own H.264 encoder. On a device whose encoder goes only to 720p, a larger video comes out at 720p (1080p as 1280×720). Where the browser has no H.264 encoder at all, a built-in encoder in the page makes it instead: each caption is drawn once as a picture (with the highlight on, once per word) and laid over the video, at no more pixels than 1280×720 (a vertical 1080×1920 video as 720×1280). The status line says so before you start.

Files the browser can't read directly (AVI, some codecs) are first turned into an H.264 MP4, then captioned. A browser that can't draw on a canvas off the page can still make the lines and save them as SRT or VTT.

What people caption

  • A 45-second talking video filmed on a phone (1080×1920), in Bold: 85 px capitals, about 24 characters per caption, so each sentence shows as a few short captions.
  • A 10-minute 1080p tutorial, the longest the free plan takes: Clean at the bottom (59 px, about 84 characters per caption) burned in, and the SRT uploaded to YouTube so viewers can also turn subtitles on in their own player.
  • An interview clip with a translator's SRT file: the speech model isn't needed; the file's tags (<i>, <b>) and positions are dropped, each cue's lines joined, and the captions drawn in the style you pick.

What goes to our server

Only the video's sound goes to Snipmill's server, when you press the button that starts the job; the line under it says so. The server deletes it once it has answered and keeps no copy. The picture isn't sent: the captions are drawn and the new video is made in your browser (usually with your device's own H.264 encoder, otherwise one built into the page). With your own SRT or VTT file, the sound isn't sent.