How-to · beginner · 2 min read
How to choose captions for short-form video
Default to 3 to 5 word phrase captions in a bold sans-serif, lower-middle of the frame above the platform UI. Word-by-word only for fast, punchy delivery.
By Sulan Zhang, co-founder of Vyra · Updated 2026-09-24 · Example: @sulansart
- Level
- beginner
- Read time
- 2 min
Most people watch muted, so captions are the main text on the video. My default is 3 to 5 word phrases, bold sans-serif, about 5 percent of frame height, sitting lower-middle above the platform buttons. Word-by-word is for fast delivery. A couple of big emphasis words per video do more than animating everything.
Why it matters
Muted viewing is the norm on all three platforms.
The method
- Pick the mode per video. Phrases for talking head, tutorials, stories, vlogs. Word-by-word for fast opinions and hype.
- One font, one weight. Vyra's default is Inter Bold for captions and Montserrat Black for titles. Don't mix three fonts.
- Size for a phone. Cap height around 4 to 6 percent of frame height on 9:16. Check it on an actual phone.
- Stay in the safe zone. Above the bottom 20 percent, inside the middle 80 percent.
- Pick 1 to 3 emphasis moments. The claim, the turn, the payoff. Make those words big and hide normal captions for that beat.
- Fix the transcript. Remove fillers, fix names, break lines where you'd pause.
Rules of thumb
| Setting | Phrase captions | Word-by-word |
|---|---|---|
| Words per caption | 3 to 5 | 1 |
| Time on screen | 0.8 to 2 s | 0.15 to 0.4 s |
| Best for | explanation, story, vlog | hype, opinion, hooks |
| Font weight | 700 | 800 to 900 |
| Animation | none | pop or instant |
| White with a soft dark outline works on almost anything. One accent color, only on emphasis words. |
Examples
- Phrase captions on a pricing talking head. "I charge by the" / "day, not the hour" / "and here is why."
- Emphasis moment. I say "I just dropped out to build something." Normal captions stop. "I just / dropped / out." stacks up big in Montserrat Black with "dropped" in the accent color, holds a second, clears, captions come back.
Do it in Vyra
- "Add phrase captions to the whole video, 3 to 5 words each, Inter Bold, white with a soft dark outline, positioned lower-middle above the TikTok UI. Remove filler words from the transcript first."
- "Switch the captions to word-by-word for the first 6 seconds only, then back to phrases. Make the word 'free' at 4 seconds an oversized emphasis word in Lemon."
- "Find the strongest claim in this transcript and turn it into a large accumulating title in Montserrat Black. Hide the ordinary captions during that moment and bring them back after."
Common mistakes
- Captions in the bottom 15 percent where the app covers them.
- A bounce on every word. It's noise after 10 seconds.
- Posting the auto transcript without fixing it.
FAQ
Burned-in or platform captions? Burn them in. Platform captions move around and are off by default for some people. Add the platform's on top if you want.
Do captions need to match exactly? Match the meaning. Cut "um" and repeats, but never change what was said.
Related
- /how-to/cut-a-talking-head
- /how-to/make-motion-graphics
- /formats/talking-head
- /prompts/captions