How to fit captions when the language has very long words

Last updated 2026-09-04 · 7 min read

What is the short version?

Some languages have words that simply do not fit a vertical frame at a normal caption size, and no amount of rewriting helps because there is no shorter way to say the thing. The answer is typographic, not editorial.

1. Go to Vizard Agent with the video and the language.
2. Name the word that is causing the problem.
3. Ask to see sizes and styles compared on real frames.

What do you need before you start?

The video and one example. A single genuinely long word — the compound noun, the inflected verb — is worth more than a description of the problem, because the size that works is decided against the worst case rather than the average one.

• The video. Captioned or not.
• The language. Russian, German, Finnish, Tamil.
• Your longest word. The one that breaks the line.
• The frame. Vertical is where this bites.
• A style you like. Or ask for a comparison.

What do you type into Vizard Agent?

Ask for the comparison rather than for a fix. There is no single right size here — it is a trade between legibility and fitting, and that trade is one you should make by looking at two frames rather than by reading a recommendation.

Prompt

Variants worth knowing:

• "Show me the styles on frames." Comparison, not a verdict.
• "Test on the longest word." Sets the worst case.
• "Break it here, not there." If you know the right split.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real vertical reel captioned in a language with long words. Steps four to eleven are one long search for a size and a style that survive the worst line in the script.

1. Transcribes every clip and pulls exact word timings.
2. Analyses the pronunciation of the key word specifically.
3. Builds the rough cut and re-transcribes it for final timings.
4. Generates animated captions and checks them on frames.
5. Optimises the word wrapping and the font size together.
6. Compares caption styles and reviews their legibility on frames.
7. Splits the captions into shorter phrases for the vertical frame.
8. Regenerates them with a dynamic highlight and re-checks.
9. Tests a smaller size specifically against the long words.
10. Finds the wrapping line in the subtitle file and inspects it.
11. Tests styles that support the script properly.
12. Compares three styles side by side on the real text.
13. Checks sizes sixteen to eighteen on the difficult words.
14. Reviews two named presets against each other.
15. Builds the final captions with the key words highlighted.
16. Reviews twelve frames across the video before rendering.
17. Renders the finished reel with the chosen treatment.

Step ten is the one that separates this from turning a size dial. Automatic wrapping breaks a line where it runs out of room, which for a long word means it breaks inside the word or pushes the whole thing to its own line — so the break point is set in the file rather than left to the renderer.

Steps twelve and thirteen are why this is a comparison and not a setting. A size that is comfortable for short words looks timid across a whole video, and the only way to choose is to see both on the same frame with the same difficult word in them.

What does the result look like?

Captions that hold every line in the script, including the worst one, at a size that still reads on a phone in a feed. Long words break where Vizard Agent decided they should break rather than wherever the renderer ran out of room, and the highlight still lands on the word being spoken.

Nothing was rewritten to achieve it. The script is what you said; only the typography changed, which matters when the words carry meaning you cannot paraphrase away.

When does this not work well?

There is a floor below which none of this helps. Past a certain size captions simply stop being readable on a phone held at arm's length, and beyond that point the honest answers Vizard Agent will give you are a different frame, fewer words on screen, or accepting two lines.

• Extremely long compounds. German and Finnish can defeat any size.
• Very narrow frames. Vertical gives you the least room.
• Scripts needing specific shaping. The typeface has to support them.
• Two lines already. A third is worse than a smaller size.
• Burnt-in captions. Changing your mind means re-rendering.

How do you fix a result that came back wrong?

Name the word and the frame. Vizard Agent keeps the subtitle file, the word timings and the style tests, so a different break point or a size change re-renders the caption layer without the cut underneath being touched.

• "It breaks inside the word." The break point set in the file.
• "Too small now." Re-tested one size up on the same word.
• "The highlight is on the wrong word." Re-timed from the transcript.
• "Try the other style." Re-rendered from the comparison set.

How does Vizard Agent compare to doing it yourself?

By hand this is a size, a render, a look, and a repeat — and the temptation is to test it on a normal line rather than on the monster in the middle of the script, which is how a caption style gets approved and then breaks at forty seconds.

Common questions

Which languages does this affect most? German, Finnish, Russian and Tamil are common cases. Tell Vizard Agent which you are in.

Can it just make the text smaller? It can, and Vizard Agent will show you what that costs on a phone.

Will it shorten my words? No. Vizard Agent treats this as typography, not as rewriting.

Can it break a word across two lines? Yes, at a sensible point. Vizard Agent sets it rather than leaving it automatic.

How do I pick a style? Look at the comparison frames. Vizard Agent renders them on your real text.

Does the highlight still work? Yes. Vizard Agent keeps the word-level timing through the restyle.

What if the typeface lacks my script? Vizard Agent tests for that and finds one that supports it.

Can I supply my own font? Yes, licence permitting. Vizard Agent uses it and tests it the same way.

Is two lines acceptable? Often yes, and Vizard Agent prefers it to shrinking. Three rarely is.

Does this apply to horizontal video? Less so. Vertical is where Vizard Agent runs out of room.

Can I see it at phone size? Yes. Ask Vizard Agent to check legibility at feed size.

What about mixed languages? Say so. The longest word may be in the other one.

Will it re-render everything? Vizard Agent redoes the caption layer only, unless they are burnt in.

Why not just shorten the sentence? Because that fixes a long sentence and not a long word, and the two problems look identical until you try it.