Most shorts captions today have one word in a different colour. Yellow is the usual choice, then red or green. It’s become so standard that a lot of editors, and most clipping tools, add it without thinking about which word it lands on.
That’s a waste, because a highlighted word is a strong signal. It’s the one thing on screen that says “this is the important part”. When it sits on the right word, the viewer gets the point of the clip even if they’re half-reading with the sound off. When it sits on the wrong word, it’s noise at best, and at worst it makes the speaker look like they’re stressing something they never stressed.
In my post on getting names and slang right, I said that choosing which words get emphasis was its own job. This is that post.
What a highlight is for
A highlight is a second layer of meaning on top of the words. The captions say what was said. The colour says what mattered.
So the question I ask for each caption chunk isn’t “which word looks good in yellow?” It’s “if the viewer only read one word here, which one should it be?” Usually there’s a clear answer, and sometimes the honest answer is “none of them”.
That framing rules out a lot of choices straight away. Articles, pronouns and filler never carry the point. Neither do most adjectives that are just there for rhythm. The word that carries the point is usually a number, a name, a verb that describes what happened, or a single word the speaker leans on with their own voice.
The words I usually highlight
These are the categories I reach for first, roughly in this order:
- Numbers that are the point. A price, a percentage, a timeline, a score. If the clip is about a trade that went from cents to dollars, the numbers are the story, so they get the colour. On business and finance clips, I wrote in the one-sentence lesson post that the number and the verb usually carry the lesson, and that holds for captions too.
- The verb that describes the move. On a fight-analysis channel I edit for, the creator made this point about our titles, and it applies just as much to captions: in a line like “he went to the body early”, the point is the decision, so the highlight belongs on “went to the body”, not on “early” or on the character’s name.
- Names, when the clip is about who. On a British NFL podcast I cut clips for, player names get the colour, because half the fun of the show is who they’re talking about. On a channel where the speaker is always the same person, their own name doesn’t need it.
- The word the speaker stresses out loud. If someone says “that was the worst deal I ever made” and leans on “worst”, the caption should lean on it too. The highlight then matches what the viewer hears, which feels natural instead of decorative.
- Short answers that land. “No.” “Never.” “Same.” When a whole point comes down to one word, that word gets the colour.
On a trades podcast I made sample shorts for, the creator’s own style was white caps with one yellow word. Going through the episode, the yellow ended up on words like “none”, “code”, “hotter”, “same” and the product numbers being compared. That list is a decent picture of what a highlight should be: the words that, read alone, still tell you what the clip is about.
The words I don’t highlight
The mistakes I see most often, in other people’s shorts and in my own early versions:
- The last word of every chunk. Some styles colour whichever word happens to come last, or the word that’s being spoken right now. That’s a karaoke effect, not emphasis. It can look good as a style, but it doesn’t tell the viewer anything, so I don’t confuse the two.
- Every emotional word. On one street-interview series I worked on, an early version had red on too many words that sounded sad. Our own review note was blunt: emphasis should highlight meaning, not every supposedly sad word. When everything is red, nothing is.
- Half of a phrase. If the meaning is in “went to the body”, colouring “the body” and leaving “went to” white splits the idea. The fight-analysis creator pointed this out on one of our early revision rounds. Highlight the whole phrase, or the one word that holds it, not a slice that happens to fit the line.
- Words the speaker didn’t mean to stress. Putting colour on a throwaway word makes the speaker sound like they meant something they didn’t. It’s a small way of putting words in someone’s mouth.
- Random picks. When I compared an AI clipping tool with my own edit, one of the clearest differences was the captions: the tool highlighted words that didn’t seem to follow any rule. Viewers may not consciously notice, but the short feels less deliberate.
One highlight per chunk, at most
My default is one coloured word or phrase per caption chunk, and many chunks with none. If a two-to-four-word chunk has nothing worth emphasising, it stays white.
That restraint is what keeps the colour meaningful. If the viewer learns that yellow only appears when something matters, they start reading the yellow words first. If yellow is on every line, they learn to ignore it.
The same goes for chunk size. On the NFL podcast, I first tried one-word caption flashes with key words in red. On minute-long clips full of crosstalk, that was tiring to watch, so we moved to two or three words per chunk. Bigger chunks also give the highlight more room to do its job, because there’s a contrast between the white words and the coloured one.
Colour roles, when a creator uses more than one
Some creators use more than one highlight colour, and then the colours need rules, not moods.
On the street-interview series, red was for shock words, green was for money, and a separate yellow was reserved for one specific speaker’s whole line, so you could tell who was talking. That works because each colour means one thing. If red sometimes means shock and sometimes just looks nice, the system stops working.
If a creator only has one highlight colour, I don’t add a second one. Matching their style matters more than my own taste.
Sometimes the right answer is no highlight at all
Plenty of good channels use no highlight colour. When I study a creator before cutting anything, I look at their own shorts first, and several of the personal-branding and biography channels I’ve looked at use plain white captions, sometimes small and lowercase, with no coloured words at all. For them, a yellow keyword would look like a different channel.
A similar lesson came from an interview podcast. The cut they ended up posting, made by another editor, used mostly small, understated captions with one large keyword per beat. My version had uniform bold caps with a coloured word in every chunk. Theirs looked more premium, and it made me rethink how loud the emphasis needs to be. Sometimes the highlight is better as a change in size or weight than as a bright colour.
So the first rule is really: copy what the creator already does. The highlight rules above are for choosing which word, inside whatever style is theirs.
A quick checklist
Before a short goes out, I read the captions with the sound off and check:
- Does every coloured word carry the point of its line?
- Is any line coloured when nothing in it matters?
- Are phrases highlighted whole, not split?
- Does the emphasis match what the speaker actually stressed?
- If there’s more than one colour, does each one mean exactly one thing?
- Does the whole thing still look like the creator’s own channel?
If the coloured words alone, read in order, give a rough summary of the clip, the highlights are doing their job.
If you’d like to see your own videos captioned this way, in your own style, I can make a free short from one of your videos. You keep it either way.