Updated October 2026 · Speech recognition and on-device transcription
Readable captions are short, well timed and placed where they do not hide the picture. Keep lines to a comfortable length, give viewers enough time to read each caption, use strong contrast, and check the result on a phone at the size most people watch. Accuracy matters as much as layout.
Captions are only useful if people can read them. A caption that appears too briefly, sits on a busy background or covers the speaker’s face is not doing its job, however accurate the words. Readability is a set of small choices about length, timing, placement and contrast, and they add up to a caption that viewers can follow without effort.
This post sets out the rules that matter most for creators, explains why each one matters and gives a simple check to run before publishing. The principles apply to live and recorded captions alike.
A caption line should be short enough to read in one glance. Long lines force the eye to travel, and viewers miss the start of the next line. Aim for a comfortable length that fits the screen width and leaves room around the edges. Break the line at a natural point, such as before a clause or a pause, rather than in the middle of a phrase.
Most captions work best in one or two lines. A caption that runs to three or four lines is hard to follow, especially on a phone where the text is smaller relative to the screen.
The reading speed is the most common cause of unreadable captions. A caption that disappears before the viewer finishes reading is frustrating. Give each caption enough time on screen to be read at a steady pace, and allow a little more for longer lines. Avoid keeping a caption on screen so long that it lags behind the speech, because the viewer then reads ahead of the speaker and loses the thread.
Check the timing by playing the video at normal speed and reading each caption as it appears. If you cannot finish a line before it vanishes, adjust the timing or shorten the text.
The bottom of the frame is the usual place for captions, but it is not always the right one. Check that the captions do not hide on-screen text, a product, a face or a chart. Some platforms place their own elements over the bottom of the frame, so check the layout on the platform itself before publishing.
Keep the position steady across the video. A caption that jumps from one side to the other makes the viewer search for the text.
Readable captions need contrast with the background. A white or light caption on a bright picture can disappear, and a dark caption on a dark scene is just as difficult. Use a background box or outline where needed, and choose a size that remains clear on a small screen. Test the captions on the busiest part of the video, not only the calmest.
Readability includes accuracy. A caption that is easy to read but wrong in a name or a number misleads the viewer. Check each caption against the audio, paying particular attention to names, numbers, dates and specialist words. Automatic captions often make the same errors repeatedly, so a single fix can improve the whole file.
Remove filler words where they add nothing, but keep the speaker’s meaning and tone. Captions are a transcript of meaning, not a perfect copy of every sound.
When more than one person speaks, identify the speaker at the start of a new speaker’s line, or use a consistent colour or label. Mark important sounds, such as laughter or music, where they change the meaning. Keep these markers brief so they do not crowd the caption.
Watch the video on a phone at the size most viewers use, with the sound off. Read every caption as it appears. Ask whether you could follow the video without the sound, and fix any caption that causes a pause. The phone check is the most useful test you can run.
A long interview needs a slightly different approach from a short, fast video. Viewers are more likely to watch in pieces, so each caption must still make sense when someone joins mid-conversation. Keep speaker labels visible at the start of each new section, and avoid captions that depend on a joke or reference from several minutes earlier.
Consider adding a short on-screen title at the start of each chapter, so viewers can orient themselves. Captions and titles together make a long conversation easier to enter at any point.
Reading speed and preference vary between viewers, and some platforms offer their own caption settings. The rules here are guidelines, not a standard that every platform enforces. Check the platform’s guidance for its own display, and keep the captions as clear as you can.
Captions help, but the sound still matters to everyone who listens. The Vocal Checker shows clipping, noise and level problems in a recording, which helps the recognition as well as the listener. The Tunedzy Go page covers cleaning a phone recording. For the export formats, see the caption export guide.
Record voice-overs with live noise removal, or clean an existing file. Runs in your browser, nothing is uploaded.
Usually one or two. Three or four lines are hard to follow, especially on a phone.
Long enough to read at a steady pace. Check by reading each caption at normal speed.
Usually at the bottom, away from on-screen text and faces. Check the platform layout.
They should match the meaning accurately. Remove filler where it adds nothing.
