Vizard Agent

How to fit translated captions that came out too long for the frame

Last updated 2026-09-04 · 7 min read

Tell Vizard Agent the captions look incomplete. It checks whether the phrase is genuinely wrapping off the frame or whether the animation is revealing it a word at a time, then rewrites the long lines as shorter translations that say the same thing — instead of shrinking the type until it fits.

What is the short version?

Almost every language takes more room than English to say the same thing. A caption style that sits neatly under an English line runs past the safe area in German, Spanish or Portuguese, and the viewer sees half a sentence.

  1. Go to Vizard Agent with the captioned version.
  2. Say the lines look cut off or incomplete.
  3. Ask it to shorten the phrasing rather than the type size.

What do you need before you start?

The captioned render and the language. What you do not need is a decision about how to fix it — whether this is a wrapping problem or a timing problem is a question with a factual answer, and guessing at it is how people end up shrinking type that was the right size all along.

What do you type into Vizard Agent?

Report what you can see and ask Vizard Agent which of the two causes it is. The two faults look completely identical on a paused frame and have entirely different fixes, so the useful version of this instruction asks for the diagnosis rather than naming one of them yourself.

Prompt

The subtitles look like they're getting cut off mid-sentence. Is that the line wrapping off the frame, or the animation revealing it late? Whichever it is, shorten the wording rather than making the text smaller.

Variants worth knowing:

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real vertical vlog whose translated captions kept arriving on screen as fragments. Steps four to six are the diagnosis, and everything after them is the editorial fix that the diagnosis pointed at.

  1. Checks the caption tool so the lines stay clear of faces and marks.
  2. Prepares concise subtitles that follow the story's order.
  3. Confirms the source is intentionally vertical, not wrongly rotated.
  4. Rebuilds the captions for the real canvas so they wrap correctly.
  5. Inspects the caption timing to see whether the apparent split is animation.
  6. Checks the subtitle engine's word timings again to be sure.
  7. Widens and tightens the layout without touching the cut.
  8. Checks one representative subtitle frame after the change.
  9. Replaces only the long phrases with concise translations.
  10. Checks the tightened treatment before delivery.
  11. Uploads the final render.

Step five is the distinction the whole page turns on. A caption that animates in word by word looks, on a paused frame, exactly like a caption that has run off the edge — and if you treat an animation as a wrapping problem you will make the text smaller for no reason at all.

Step nine is the fix that keeps the video readable. Shrinking type to fit a long translation makes every caption harder to read, including the ones that were fine; rewording the three lines that overflow leaves the rest untouched.

What does the result look like?

Captions that read as complete thoughts, at the size they were designed to be, sitting comfortably inside the safe area of the frame. The handful of lines that overflowed now say the same thing in fewer words rather than in smaller ones, and every other caption is exactly as it was.

Where a term could not be shortened — a name, a product, a place — it survives intact, and the trimming happens around it.

When does this not work well?

There is a floor below which shortening a translation stops being accurate and starts being a different claim. Some ideas need the words they need, and some scripts simply cannot be delivered in the time the original speaker took to say them in English.

How do you fix a result that came back wrong?

Say which line is wrong and what you think it lost. Vizard Agent keeps the word timings, the caption layout and the full untrimmed translation, so restoring a clause or re-shortening a different line is an edit to those lines alone rather than a new caption pass.

How does Vizard Agent compare to doing it yourself?

By hand the reflex is to reduce the font size, because it is one setting and it works immediately. It also makes the whole video slightly harder to read to solve a problem that exists on four lines out of sixty.

By hand Vizard Agent
Diagnosis Assume it is wrapping Wrapping and timing told apart
The fix Smaller type Shorter wording on the lines that overflow
The other captions Also shrunk Untouched
Protected terms Trimmed by accident Kept deliberately
Checking On a paused frame On the rendered output

Common questions

Why are translated captions longer? Most languages need more characters than English for the same meaning.

Will shortening change what it says? Vizard Agent keeps the meaning and cuts the words. Say if a line is sensitive.

Can it just use a smaller font? It can, and Vizard Agent will explain why that is usually the worse trade.

How do I know it is wrapping and not animation? Ask. Vizard Agent checks the word timings rather than guessing.

Can I protect certain words? Yes. Name them and Vizard Agent trims around them.

Does it handle two-line captions? Yes, and it will tell you when a third line is the wrong answer.

What about right-to-left languages? Handled, though the wrapping rules differ. Say which language.

Can it keep the same caption style? Yes. Vizard Agent changes the words, not the treatment.

Will the timings still match? Yes. Vizard Agent re-derives them for the shortened lines.

What is the safe area? The part of the frame the platform's buttons do not cover. Vizard Agent works inside it.

Can it do this before the render? Yes. Tell Vizard Agent the target language up front and it writes to length.

Does it help to see the original English? Yes. Vizard Agent keeps both so you can check what a shortened line dropped.

Can it check every line? Yes. Ask Vizard Agent to scan rather than fixing only what you spotted.

Does this work for burnt-in captions? Yes, though Vizard Agent has to redo the render for them.

Can it show me the before and after? Yes, on real frames rather than in a text file.

Why not translate more literally? Because a literal caption nobody can finish reading conveys less than a shorter one that fits.