How to stop your captions breaking words across two lines

Last updated 2026-09-10 · 8 min read

What is the short version?

A caption renderer wraps text to fit a width, and when a word does not fit it breaks the word. Asking for a different style does not change that. The fix is to decide the breaks before the renderer sees the text.

1. Go to Vizard Agent with the clips.
2. Say every word must stay whole.
3. Say how many lines you want.

What do you need before you start?

The clips and a short description of the caption style you want. The important part is the rule about words, which is easy to state and very easy to leave out — most people ask for "two lines" and assume whole words come with it.

• The clips needing captions.
• The rule. Words never split across lines.
• How many lines. Two is the usual answer.
• The style. Classic and plain, or animated.
• The size. Bigger text breaks more often.

What do you type into Vizard Agent?

State the rule as a rule. "Make the captions neater" is a preference and will be interpreted; "break lines at word boundaries, never inside a word" is a condition every line in the file can be checked against afterwards.

Prompt

Variants worth knowing:

• "Never inside a word." The rule, stated plainly.
• "At word or phrase boundaries." Where breaks should fall.
• "Two lines." Otherwise you may get three.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real set of vertical clips whose captions kept splitting words. The interesting part is that it takes three attempts, and the third one changes the approach rather than the settings.

1. Transcribes the recording with word-level timings.
2. Prepares the captions and renders the first set of clips.
3. Reviews the frames for framing, titles and caption placement.
4. Fixes the wrapping along with the pauses and sentence endings.
5. Regenerates the captions in the plain two-line style.
6. Extracts a contact sheet to check the revised style.
7. Reviews the style and the line wrapping across the clips.
8. Eliminates the remaining mid-word wraps and rebuilds.
9. Verifies every subtitle line keeps its words intact.
10. Builds the subtitle files with the breaks already decided.
11. Confirms they wrap only at word boundaries.
12. Increases the size while preserving the whole-word wrapping.

Step ten is the change of approach. Up to that point the renderer is being asked to wrap better, and it wraps to fit a width because that is what wrapping means; writing the files with the breaks already in them removes the decision from the renderer entirely.

Step one earns its place here. Whole-word wrapping needs to know where the words are, and a transcript with timings on every word gives Vizard Agent the boundaries directly rather than leaving it to infer them from spacing in a block of text.

Step twelve is the check that catches the relapse. Making the text bigger makes every line wider, so a caption set that wrapped perfectly at one size will start splitting words at the next, and the rule has to be re-verified after any size change.

What does the result look like?

Captions in a plain two-line style where every single line ends at the end of a word, set at a size you can comfortably read, and holding across every clip Vizard Agent delivered rather than across most of them.

Nobody notices captions that wrap properly, which is the entire objective.

When does this not work well?

There are cases where no break position helps at all. A word can simply be longer than the line. In languages that build long compound words, or with a very large type size on a narrow vertical frame, the word will not fit however Vizard Agent breaks around it, and the real choice is between a smaller size and a wider layout.

• Words longer than a line. Something has to give.
• Very large type. Fewer words fit before a break.
• Narrow vertical frames. Less width to work with.
• Compound languages. Long words are normal there.
• Animated word-by-word styles. They wrap differently.

How do you fix a result that came back wrong?

Point at the clip and quote the line that is wrong in it. Vizard Agent keeps the subtitle files it wrote for every clip, so a fix moves a single line break in one of them rather than regenerating all the captions from the audio all over again.

• "This one still splits a word." That break moved to the boundary.
• "Three lines here." Re-wrapped to two, or the size reduced.
• "Too small now." Enlarged, with the rule re-verified.
• "The phrasing reads oddly." Broken at phrase boundaries instead.

How does Vizard Agent compare to doing it yourself?

By hand this is treated as a styling problem, so the style gets changed and the words keep breaking. Checking every line is also the part that gets skipped — a spot check finds the obvious ones and leaves the two in the middle of clip four.

Common questions

Why does it break words at all? Because wrapping fits a width. Vizard Agent decides the breaks instead.

Can I have exactly two lines? Yes. Say so and Vizard Agent works to it.

What if a word is too long? Vizard Agent will reduce the size or tell you the layout is too narrow.

Will it break at phrases? Yes, if you ask. Vizard Agent finds it reads better than breaking anywhere.

Does this work with animated captions? Yes, though Vizard Agent handles their wrapping differently.

Can I change the size afterwards? Yes, and Vizard Agent re-verifies the wrapping when you do.

Will it check every clip? Yes. Ask Vizard Agent to verify every line, not a sample.

Can I get the subtitle files? Yes. Vizard Agent exports them alongside the clips.

What about other languages? Same rule. Vizard Agent applies it per language.

Does the timing change? No. Vizard Agent changes only where the lines fall.

Can it match a caption style I have seen? Yes, and the whole-word rule still applies on top.

What about punctuation at a break? Vizard Agent keeps it with the word it belongs to.

Will hyphenated words split? Only at the hyphen, and only if you allow Vizard Agent to.

Why not just use a smaller font? Because a smaller font hides the fault at that one size, and it comes straight back the moment anybody decides the captions need to be bigger.