Readable caption styles for vertical video

What size should a caption be on a vertical video?

Start from the frame, not from a font menu. On a 1080x1920 frame we burned the same 27-character caption into a real clip at seven font sizes, then scaled each frame to a 390 CSS px wide phone view and measured the height of the letters and the width of the line.

Two constraints fight each other. Bigger letters are easier to read, but a longer line stops fitting across the frame. The table below is what we measured for the sentence "TAP TO SEE THE FULL ROUTINE" in a bold sans serif.

Our measurements, October 26, 2026. We rendered the same caption with ffmpeg 6.1.1 in DejaVu Sans Bold on a 1080x1920 frame, measured the letter pixels in the rendered file, and scaled the frame to 390 CSS px wide to get the phone column. The letter height is the cap height of the capital letters, so a sentence in mixed case is slightly taller. From 64 px up the line ran off the frame, so the character counts for those rows are scaled from the 56 px width, not measured.

Read the table this way: 48 px is the largest size where a 27-character line still fits inside 80% of the frame width, and 13.4 CSS px is the height those letters take on a phone screen. If your caption is a short phrase, size it by the width instead: our sample caption at 32 px still fills only 52% of the frame, so a three-word line can be twice as large as a full sentence. The 80% limit is our working margin from the frame edges, not a platform rule, and a font wider than DejaVu Sans Bold reaches it at a smaller size.

Which caption style should I use?

The style matters more than the font. We rendered seven caption styles onto one frame from each of two real CC0 clips, then compared the brightness of every letter against the pixels it actually touches, using the contrast formula from WCAG 2.2.

The left column is a dark garden clip. The right column is a snowy street in daylight, which is the harder case: the brightest patch under the caption was almost as light as the text itself.

Our measurements, October 26, 2026, on the rendered pixels of one frame from each clip, with a 48 px caption and the WCAG 2.2 relative luminance formula. The figure is the same caption three times on the same snowy frame.

WCAG 2.2 asks for 4.5:1 for normal text and 3:1 for large-scale text. It is a standard for web content, not for video, so we use it as a yardstick rather than a rule. Each number is the contrast that the worst tenth of the letters reached. On the snow clip one letter in ten of plain white text sat below 4.5:1, and the brightest pixel under the caption was within 1.15:1 of white, so a white letter landing on it all but disappears.

The outline is the reason it wins: a black stroke sits between the white letters and whatever the camera saw, so the letters never touch the snow. A shadow cannot do that, because the shadow only touches one side of each letter and the other side still sits on the footage.

How long should a caption stay on screen?

A caption that is gone before it is read is a caption that was not written. The speed number to use is characters per second: the length of the line, spaces included, divided by the seconds it stays on screen.

Netflix publishes two limits for English and they do not match, so we used the stricter one. Its English (USA) Timed Text Style Guide allows up to 20 characters per second for adult programs; its Subtitle Templates guide caps adult English templates at 17 characters per second. We used 17.

Our calculation, October 26, 2026, from the caption strings above and the durations in the table. Netflix also requires a minimum of five sixths of a second per subtitle event and a maximum of seven seconds, and two lines of text at most.

Practical rule: seconds on screen = characters divided by 17, then round up. A 27-character caption needs 1.6 seconds, and a 34-character caption needs 2.0 seconds. If the line does not fit that time, split it or cut words rather than speeding it up.

What do the published caption size rules say?

Creator guides disagree with each other, and none of them is a platform rule. We checked the TikTok and YouTube help pages: TikTok says creator captions let you change the font style and the color, and YouTube describes manual captions, automatic captions and caption files. Neither publishes a size.

What the guides do publish are percentages of the frame height. They are worth reading together, because the spread between them is the whole problem.

Third-party creator guides, read October 26, 2026. The pixel column is our arithmetic on a 1920 px tall frame; the rules themselves are community advice, not platform documentation. The largest published number is more than twice the smallest.

Part of the disagreement is units. These percentages are quoted against font size, while our table measures the height of the letters, and the two are not the same thing. Our own answer is narrower because it starts from a real constraint: 48 px is the largest size that keeps a full sentence inside 80% of the frame width.

What values should you enter in your editor?

Set the values, not the menu names. Editors express caption size differently, so match what appears on the frame rather than a number in a field.

We did not test inside CapCut, Premiere Pro or DaVinci Resolve. The numbers below are what we measured on the rendered pixels; the conversions from pixels to points are ours at 1080p, where one pixel is 0.75 points.

Our measurements and our arithmetic, October 26, 2026. For the step-by-step caption workflow in each app, see our CapCut caption guide and our subtitle file guide, both linked below.

Where should the caption sit in the frame?

Size and contrast decide whether a caption can be read. Position decides whether it survives the platform interface, which covers the bottom of a vertical video with the caption, the name and the sound label, and the right side with the like, comment and share buttons.

Keep the caption centered horizontally and high enough to clear that layer. Our safe zone templates show the published margins for TikTok, Instagram Reels and YouTube Shorts, and the safe zone calculator will place a caption box for you.

Limits and what we did not test

• We did not test on a physical phone. We scaled every frame to a 390 CSS px wide phone view and report the letter height in CSS pixels as our proxy for physical size. CSS pixels are not inches, so treat the phone column as a comparison between our sizes, not a measurement of a real screen.
• We did not test inside CapCut, Premiere Pro or DaVinci Resolve. Every number here comes from rendered pixels and from the published rules in the sources below.
• We tested two CC0 clips, one dark and one bright. Footage that moves behind the caption, such as a busy street or water, can be worse than either, because the background changes while the text is on screen.
• We did not test color vision, screen readers or the accessibility of a caption track. WCAG 2.2 is a web standard and we use its contrast formula as a yardstick only.
• We did not test captions in other languages. Netflix publishes different reading speed limits per language, and those are the numbers to look up for a language other than English.