Word by Word Captions: How Timings Survive Every Cut
Word by word captions come from a transcript that knows when each word starts and ends. The editor groups the words into short lines, shows each line while it’s spoken, and highlights the current word as you hear it. The hard part comes after you edit: every cut and speed change moves the words, so the timings have to be recalculated on the final video, not the raw clip.
What is a word timing?
A normal subtitle file says “this sentence, from 3.2 to 5.8 seconds”. A word timing goes finer: “this” from 3.20 to 3.41, “sentence” from 3.45 to 3.90. Speech-to-text tools can return those timings for every word, and generated voice-overs come with them built in. Word by word captions are impossible without them; with them, they’re mostly arithmetic.
Why do captions drift after you edit?
Because the words were timed against the original clip. Three edits break them:
| Edit | What happens to the words | The fix |
|---|---|---|
| A cut | Everything after it moves earlier | Subtract the removed time from every later word |
| A clip used out of order | Its words move to wherever the clip now sits | Add the clip’s new start on the final video |
| A speed change | Words get closer together (faster) or further apart (slower) | Divide each word’s offset by the speed |
| A transition | Two clips overlap for a moment | Shift the second clip’s words by the overlap |
Do this by hand and you’ll be nudging captions for an hour. An editor that knows where each clip came from can do it exactly. Take the word’s time in the source and subtract the clip’s in point. Divide by the speed. Add the clip’s start on the final video.
Pop, karaoke or highlight: what’s the difference?
All three show a short line and mark the word being spoken. They mark it differently:
- Highlight: the current word changes color. Clean and readable; works for almost anything.
- Pop: the current word changes color and grows a little. More energy, suits fast reels and talking heads.
- Karaoke: the line fills in as it’s spoken, left to right. Suits music, storytelling and slower voices.
Our post on captions for reels covers which style, size and position to pick. This one is about how they’re made.
How many words per line?
Two to four for short vertical video, so the type can be big. Long-form 16:9 video can carry more, four to seven, in one or two lines at the bottom, because the frame is wider and viewers sit further from the screen. Break lines at natural pauses: after a full stop or a gap in the speech. A line that runs across a pause looks wrong, because the highlight sits still on the last word.
What about languages and accents?
Transcripts can detect the language. Word timings work the same in any language with spaces between words. Two things to check: a font that has every letter of the language (accents, special characters), and names or jargon the transcript might spell wrong. Fix those in the text; the timings stay. Mixed-language speech works too, but check the spelling of the second language before you render.
How does Cut Express make word by word captions?
Cut Express is a video editor you drive from Claude.
- Word timings. For a clip with speech, Claude asks for a transcript (0.1 credit per minute of audio, once per file). A voice-over made in Cut Express already has its timings.
- Mapping. When Claude renders, Cut Express maps every word through the timeline for you, across every cut and speed change. Claude never converts a timestamp by hand.
- Styles. Six presets (pop, karaoke, boxed, bold, clean, minimal), each changeable: font, size, colors, highlight color, outline, a plate behind the line, position, case, words per line, one or two lines.
- Burn-in. The captions are drawn into the video, so they show on every platform without a subtitle file.
Captions themselves spend no credits; only the transcript does. The features section lists the rest of what a timeline holds.
What prompt gets the captions you want?
“Word by word captions in the pop style, three words per line, uppercase, yellow highlight, at 72% height, never over my face.” Then notes: “bigger”, “no highlight”, “two lines”. Each note renders a new version.
If your video removes pauses and false starts first, captions follow the cut automatically. See removing silences.
FAQ
Can I write the captions myself? Yes. Give the lines and their times; they’re shown as written, with the words spread over each line.
Do captions follow a voice-over track? Yes. Words from a generated voice or a transcribed voice track land where the track sits on the video.
Will a 2x speed-up break them? No. Speed is part of the mapping, so the words land where they’re heard.
Do I get an SRT file? Captions are burnt into the MP4. A separate subtitle file isn’t part of the export.
More answers in the FAQ and on the blog. Plans on pricing.
Caption your next video with Cut Express
Cut Express is the video editor you drive from Claude: upload the clip, name the style, get the MP4 with captions that land on every word. See the plans.