Most creators turn captions on, keep the default style and post. The default is built to look good in a demo, not to be read on a phone by someone half paying attention. These are the settings that make captions readable, with the reason for each so you can adjust them.
Why every short video needs captions#
- Muted viewing. Many people scroll with the sound off. A claim that exists only in audio does not reach them.
- Comprehension with sound on. Reading and hearing the same words at once survives distraction better than either alone.
- The first frame works as a still. A frame with a sentence on it says something as a thumbnail. A face mid-blink does not.
Position: inside the safe zone#
Do not centre captions vertically and do not put them at the very bottom. TikTok, Reels and Shorts all draw their own interface over the bottom and right edge: captions, handle, sound strip and buttons.
A safe band for your text sits about 70 to 80 percent of the way down the frame. Keep text between 8 and 92 percent across. Set this once in your template and never drag it by eye again, or one day your call to action ends up behind a share button.
Line length: two lines, short ones#
- Two lines at most. Never three.
- About 20 to 28 characters per line. Longer lines force side-to-side eye movement on a phone.
- Four to seven words on screen at once for block captions. Fewer turns text into flicker. More turns it into a paragraph.
If your tool breaks lines automatically, check that it never splits a word.
Reading speed#
A caption must stay on screen long enough to read. Broadcast subtitle guides cap reading speed somewhere around 15 to 20 characters per second, and a distracted phone viewer reads slower. Normal speech runs around 150 words a minute, so fast speech can flash a line away before anyone reads it.
When the script is dense, caption the claim, not every word. TV subtitles have condensed speech for decades for this reason.
If the goal is accessibility, this reverses: you want a full, accurate transcript. Know which job you are doing.
Contrast against a moving background#
White text that reads well over a dark shot disappears when the clip cuts to a bright window. Three defences, from lightest to heaviest:
- Stroke. 3 to 4 pixels of near-black around white text at 1080 wide. Usually enough.
- Drop shadow. A few pixels down, a soft blur, black at partial opacity. Reads as depth.
- Background plate. A dark box behind the text with some padding. Always legible, and looks like a caption rather than part of the shot.
Use one or two, not all three. Stroke plus shadow survives the most backgrounds.
Type size, tested at arm’s length#
At 1080 by 1920, caption text below about 44 points starts to fail. Most short-form captions sit between 48 and 64 points. Hook text can go much larger, 72 to 96, because one big sentence reads in the first half second. Text-based videos for TikTok build a whole format on that.
Weight matters as much as size. A bold caption at a smaller size often reads better than a regular one that is larger. Test on a real phone held where you normally hold it, not at your desk.
Block or karaoke#
- Block: the whole line appears and holds. Best for explainers, objections, numbers and anything a viewer may re-read.
- Karaoke: words highlight as they are spoken. Best for pace and energy, and when the audio is the content.
Pick one per format, not per video. A channel where the style changes every post looks unfinished.
Skip the word-by-word bounce#
Most caption tools default to one word at a time, scaling in, sometimes rotating. It looks lively in a demo. In a feed it makes text impossible to skim and turns the caption into motion that fights the video. Use it on three or four hook words at most. The same goes for entrance animations: pick one for the hook and none for the rest.
What not to caption#
- Filler words such as “so, um, basically”.
- Sound effects, unless you are doing full accessibility captions properly.
- The brand name every time it is said.
Keep other text out of the caption zone. A caption and a call to action in the same band will collide on a long line.
Burned-in versus platform auto-captions#
Platform auto-captions are added after upload, can be turned off by viewers, and are styled and placed by the platform. Good for accessibility, no use for design.
Burned-in captions are part of the rendered video. You control position, style and timing, and they survive downloads and reposts.
For brand content, burn them in and leave platform captions on as well where you can. Keep your burned-in captions inside the safe zone so they do not stack on top of the platform’s.
The text you post under the video is a separate limit: TikTok and Instagram allow 2,200 characters. The caption length checker tests yours against each platform.
The template in numbers#
- Position: about 78 percent down the frame.
- Horizontal bounds: 8 to 92 percent.
- Lines: 2 at most, 20 to 28 characters each.
- Size: 48 to 64 points at 1080 wide, bold weight.
- Line height: 1.05 to 1.15.
- Stroke: 3 pixels near-black, plus a soft shadow.
- Style: block for explanation, karaoke for pace. One per format.
- Animation: none, except on the hook.
Set these once and the decision leaves your week. Looptake adds word-timed animated captions to every video it makes, placed in the safe zone and styled from your brand kit, so the same settings hold across hundreds of posts. If you post one video to several platforms, repurposing one video for TikTok, Reels and Shorts covers the rest of the checklist.