🎓 Journalism Training
Courses › Video and Reels for Social Media

Lesson 3 of 6 · about 12 minutes

Captions burned in

You'll be able to

  • Explain why reels need captions burned into the picture
  • Set caption size, color and line length that read on a phone
  • Break a transcript into short caption chunks of two lines or fewer
  • Place captions above the bottom overlay and inside the safe area

Most people meet your reel on mute. On a bus, in a waiting room, in bed next to someone sleeping. If the story lives only in the audio, those people get a stranger moving their mouth.

The best-known figure comes from a 2019 survey by Verizon Media and the agency Publicis Media: of 5,616 US adults aged 18 to 54, 69% said they watch video with the sound off in public places, and 25% said they do so even in private. The survey is older than most of the apps' current designs, and it measured video in general rather than reels. Treat it as the best public number available and hold it loosely. You will see bigger figures, like "85% of Facebook video is watched muted," repeated everywhere. That one traces back to a May 2016 Digiday article in which a couple of publishers reported their own Facebook numbers. It was a handful of companies describing their own audiences in 2016, before Reels existed. Do not cite it as a general fact.

The craft answer does not depend on the exact number. A big share of your audience cannot hear you at the moment they decide whether to stay. So the words go on the picture.

Burned in, every time

Burned-in captions (also called open captions) are drawn into the video pixels. Closed captions are a separate track the viewer can switch on. For reels, burn them in.

Closed captions are good practice for long videos on YouTube, and you should upload a caption file there too. But in a short-video feed, almost no one turns captions on. Each app handles caption files differently, some not at all, and the auto captions they offer are not always switched on. TikTok, to its credit, added auto captions in 2021 partly for deaf and hard-of-hearing users. Those still depend on the speech recognition getting your words right. Burned-in text shows the same on every app, every phone and every repost.

Burned-in captions are part of the picture. You cannot fix a typo after export. Proofread every caption before you render.

Size, font and color

A reel is watched on a screen about the size of a playing card. Text that looks large on your monitor looks timid there.

  • Size: on a 1080x1920 frame, caption letters should be roughly 60 to 80 pixels tall for capital letters, and never under about 48. If you have to squint at the export on your phone at arm's length, go bigger.
  • Font: a bold or semibold sans-serif. Thin, script and condensed fonts fall apart at phone size.
  • Color: white or pale yellow letters with a dark stroke (outline) of about 4 to 8 pixels, or a semi-opaque dark box behind each line. One of the two is required. Plain white text with no outline vanishes over a bright sky or a white wall.
  • Case: normal sentence case reads faster than all capitals. Save capitals for one word you want to land.

A single emphasis color is fine. Many creators color one key word per caption, like a dollar amount, in yellow. Do that sparingly. When every other word is colored, nothing stands out and the screen looks like a ransom note.

Two lines, short chunks

A caption chunk is what shows at once. Keep it to two lines at most, and one line is often better. Each line should be about 25 to 32 characters at the sizes above, which is four to six words.

Break lines at natural pauses: after a comma, before "and" or "but," between a subject and a long phrase. Never split a name, a number and its unit, or a word from the word it belongs to.

Example: chunking one sentence

Spoken: "The Ridgeline school board voted four to one last night to close Elm Street Elementary at the end of this school year."

Bad chunking, one huge block:

The Ridgeline school board voted four to one last night to close Elm Street Elementary at the end of this school year.

Better, four chunks shown one after another:

  1. The Ridgeline school board
    voted 4 to 1 last night
  2. to close Elm Street Elementary
  3. at the end
    of this school year.

Each chunk is short enough to read in the time it takes to say it. TikTok's ad guidance suggests 5 to 10 words per second of on-screen text, and that is about the top speed for most readers.

Word-by-word "karaoke" captions, where each word pops on as it is spoken, are popular. They work for fast, punchy talk. For news, where the viewer may want to reread a number, phrase-length chunks that stay up a little longer are kinder.

Placement

TV subtitles sit at the bottom of the screen. Do not copy that on a reel. The bottom of a reel is where the app puts the username, the post caption and the audio label, and your burned-in captions will be buried under them.

Use the safe area from the framing lesson. Keep captions above the bottom 35% of the frame, which on 1080x1920 means the bottom of your caption box should sit no lower than about 1,250 pixels from the top. Center them horizontally and keep them clear of the button column on the right.

A good default spot is just below the speaker's chin. The viewer's eyes are already on the face, so the words are a short glance away. Keep the position fixed for the whole reel. Captions that jump around the frame make the viewer hunt.

And keep the text from taking over. Instagram's creator guidance says recommended content should not have text covering the majority of the screen (as of October 2026). Two lines of captions and a short hook are nowhere near that. A full paragraph pasted over the video is.

Auto captions are a first draft

Speech recognition is good and fast and wrong just often enough to hurt you. It mangles names, local place names and numbers. "Fifteen" becomes "fifty." "The levy" becomes "the levee." In a news reel, a wrong word in a caption is a misquote you published.

Run auto captions to save typing, then read every chunk against the audio. Check every name, every number and every negative ("did" against "didn't" is the classic). Leave out filler like "um" and false starts. That is normal caption practice. Never change what someone meant.

Try it

  1. Take a 30 to 45 second clip of someone talking. Write out exactly what they say.
  2. Break it into caption chunks of one or two lines, about 25 to 32 characters per line, with breaks at natural pauses.
  3. In VideoStudio or AVPrep3, add the captions as text on a 1080x1920 project: bold sans-serif, white with a dark outline or dark box, placed under the speaker's chin and above the bottom 35%.
  4. Export, watch it on your phone with the sound off at arm's length, and fix anything you could not read in one pass.

The AV craft chapter Text people can read goes further on type size and contrast. If the reel is a music video, Lyric Bench is built for timed lyric text.

Checklist

  • Captions burned into the picture
  • Bold sans-serif, with outline or dark box
  • Two lines at most per chunk, broken at natural pauses
  • Placed above the bottom overlay, same spot for the whole reel
  • Every name, number and negative checked against the audio

Check yourself

1. In the 2019 Verizon Media and Publicis Media survey of 5,616 US adults, what share said they watch video with the sound off in public?

2. What does 'burned in' mean for captions?

3. What is the most a single caption should show at once on a reel?

4. Where should burned-in captions sit on a 1080x1920 frame?

5. Which caption style is easiest to read on a phone?

6. Your auto-caption tool wrote 'the boarded proved the levee' for 'the board approved the levy.' What should you do?

Terms in this lesson

Burned-in captions
Captions drawn into the video picture, so they show everywhere with no setting. Also called open captions.
Closed captions
Captions stored as a separate track or file that the viewer can turn on or off.
Caption chunk
One block of caption text shown at a time, usually a short phrase of one or two lines.
Auto captions
Captions made automatically by speech recognition. Useful as a first draft, never as a final one.
Stroke
An outline drawn around each letter so text stays readable over light and dark backgrounds.

Full glossary

Sources