How to add subtitles to a video (and make them readable)
Line length, reading speed, line breaks and placement: the rules professional subtitlers follow, with examples, and how to apply them to your own videos.
Getting the words right is only half of subtitling. The other half is making sure someone can actually read them before they disappear, without squinting, and without the text covering the one thing they needed to see.
Plenty of people watch with the sound off: on a train, in an open-plan office, next to a sleeping baby. Subtitles are also how deaf and hard-of-hearing viewers follow your video at all, and they help anyone watching in their second language. So it’s worth doing them properly.
This guide covers the practical rules. Most of them come from broadcast and streaming style guides, and they hold up just as well for a product demo or a lesson.
Subtitles, captions, open, closed: a quick glossary
People use these words loosely, so here is what they usually mean.
- Subtitles assume the viewer can hear but may not understand the language. They carry the dialogue, often translated.
- Captions assume the viewer can’t hear the audio. They include dialogue plus sound cues like [phone ringing] or [laughter], and they name the speaker when it isn’t obvious. In the UK, “subtitles” is commonly used for both.
- Closed captions live in a separate file (such as SRT or VTT) that the viewer can switch on and off.
- Open or burned-in captions are part of the picture. Everyone sees them, and they look the same on every platform.
For social video, burned-in captions are the norm because you control how they look and they show up even where the player has no caption button. For long videos on YouTube, many creators do both: styled captions in the video for key moments, and a caption track uploaded in YouTube Studio for the full transcript.
Step 1: Get an accurate transcript
You have three options. Type it yourself, import a subtitle file you already have, or let speech recognition do a first pass.
In Glintia’s subtitle generator you can do any of the three: transcribe a clip on the timeline with AI, write captions by hand, or import an SRT or VTT file. Transcription costs 0.25 AI credits per minute of audio, so a five-minute video uses about 1.25 credits.
Whichever route you take, read the transcript while listening. Automatic transcription is good now, but it still trips on the same things:
- Names of people, places and products
- Numbers, prices and dates
- Industry terms and acronyms
- Two people talking over each other
Fix those first. A misspelled product name in a caption looks worse than having no caption.
Step 2: Break the text into readable chunks
This is where most DIY subtitles go wrong. The transcript is right, but it’s chopped into lines that are too long, too fast, or split in odd places.
Netflix’s English timed text style guide is a good public reference. Its main rules:
- 42 characters per line, at most.
- Two lines per subtitle, at most. Keep it to one line when it fits.
- Up to 20 characters per second of reading speed for adult programs, and up to 17 for children’s programs.
That reading speed rule is the one people break most. Take this sentence:
If you can’t find the link, check the email we sent you this morning.
It’s 69 characters. At 20 characters per second, it needs to stay on screen for about three and a half seconds. If your speaker says it in two seconds, the caption has to be shortened or split across two subtitles, or viewers will miss the end of it.
Where to break a line
The same guide recommends breaking after punctuation, or before a conjunction or preposition. It also says to keep certain words together: an article with its noun, a first name with a last name, a verb with its subject pronoun. When you have a choice, make the bottom line the longer one.
Here’s the same sentence broken two ways.
Awkward:
If you can’t find the link, check the
email we sent you this morning.
Better:
If you can’t find the link,
check the email we sent you this morning.
The second version breaks at the comma, keeps “the email” together, and puts the longer line at the bottom where the eye lands. It reads in one glance.
What about one-word-at-a-time captions?
The big, bold, two-or-three-words-at-a-time captions you see on TikTok and Reels follow different rules. They’re designed to grab attention in a fast feed, not to carry a long explanation. They work well for hooks and short clips. For a three-minute tutorial, they get tiring, and normal sentence-based subtitles are kinder to the viewer.
Glintia’s karaoke mode sits in between: the whole line appears, and each word highlights as it’s spoken. It keeps the line readable while still giving the eye something to follow.
Step 3: Style for contrast, not decoration
A caption’s first job is to be legible on top of whatever is behind it. White text on a bright sky disappears. So does yellow text on a beige wall.
The fixes are simple:
- Add an outline (stroke) or a soft shadow around the letters.
- Or put the text on a semi-transparent box, the classic “movie subtitle” look.
- Use a clean, heavy-enough font. Thin decorative fonts break up at small sizes.
- Make it big enough to read on a phone. Check it on an actual phone before you export.
Glintia has 51 caption presets, from Classic White and Cinema Box for long-form work to bolder social styles, plus controls for stroke, glow and background opacity. Pick one that suits the video, apply it to every subtitle at once, and only override individual lines when you have a reason to.
Step 4: Put captions where they won’t block anything
The bottom center of the frame is the default for a reason. Viewers expect to look there.
Move the captions up when something important is already at the bottom: a name tag, a product shot, on-screen instructions, or a screen recording where the useful part is low in the frame. Glintia lets you position subtitles at the top, center, bottom or a custom spot.
Vertical video needs extra care. TikTok, Reels and Shorts put the account name, the description and the like buttons along the bottom and right edge. Keep captions above that area, or the platform will draw right over them. And try not to place a caption over someone’s mouth or face for long stretches.
Step 5: Check the timing
Good caption timing is mostly invisible. A few things to watch for:
- The caption should appear when the speaker starts talking, not a beat later.
- It shouldn’t vanish before a slow reader could finish it. The characters-per-second rule above is your guide.
- Avoid flashing a caption on screen for a fraction of a second. Merge it with the next one.
- Try not to let a caption hang across a hard cut to a completely different scene.
In Glintia you can split a subtitle in two, merge it with the next one, and drag its timing on the timeline, so fixing a cue that’s too fast takes a few seconds.
Step 6: Translate if your audience needs it
Translated subtitles reach viewers who would otherwise scroll past. Glintia can translate a caption track into any of 20 languages with AI and keep the timing, for 1 credit per translation. You can show the original, the translation, or both at once, which is useful for bilingual audiences and language learners.
Two things to check after translating:
- Line length. Translations often run longer than the English. Recheck that lines still fit and that the reading speed is still comfortable.
- Terms that shouldn’t be translated. Product names, button labels that match an English-only interface, and people’s names.
If you can, have a native speaker skim the result. Machine translation is a great first draft and a risky final one.
One more thing: filter what you didn’t script
If you’re subtitling interviews, customer videos or anything unscripted, someone may say something you’d rather not put on screen, like a swear word or a phone number. Glintia’s content filter scans the transcript for profanity, sensitive details such as phone numbers and email addresses, and words you add yourself. It can then mask the caption and mute or bleep the audio.
A subtitle checklist
- Transcript proofread against the audio, especially names and numbers
- No line over 42 characters, no subtitle over two lines
- Reading speed under about 20 characters per second
- Lines broken at natural phrase boundaries, bottom line longer
- Strong contrast: outline, shadow or background box
- Nothing important hidden behind the text, and clear of platform buttons on vertical video
- Translated tracks rechecked for length
Subtitles pair naturally with narration. If you’re generating a voiceover too, our guide to making text-to-speech sound natural covers the audio side.
Try Glintia free for 3 days and caption your next video in the same editor you use to make it.