How to add a translated subtitle under the original line without a box

Last updated 2026-09-10 · 8 min read

What is the short version?

A translated subtitle added on top of a video that already has one ends up either over the original line or in the middle of the frame, and the black plate behind it covers the faces. Both problems come from placing it without measuring what is already there.

1. Go to Vizard Agent with the video.
2. Say the translation goes under the existing line.
3. Say no black box behind it.

What do you need before you start?

The video and a decision about the plate behind the text. Most subtitle tools add a background by default because it guarantees legibility over anything at all; on a video where the faces are the whole point, that guarantee costs Vizard Agent more than it is worth.

• The video with its existing subtitles burnt in.
• The target language.
• Where the translation sits. Under the original, usually.
• Whether a plate is allowed. Usually not.
• The look you want. Bold and flat reads best.

What do you type into Vizard Agent?

Say what the box is doing. "Remove the background" sounds like a style preference; "it is covering their faces" is a reason, and it also tells Vizard Agent what the subtitle has to avoid once the plate is gone.

Prompt

Variants worth knowing:

• "Underneath the English line." A position relative to what exists.
• "Without the black box." And say why.
• "Bold so it still reads." What replaces the plate.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real translated video whose original subtitles were burnt into the picture. Almost every step is a measurement or a close-up check, because none of this can be judged at normal size.

1. Tests the covering layer alone on a source frame to find the original line's coordinates.
2. Separates the covering test from the translation to see what is really there.
3. Measures the covering layer's bounds in the exported frame.
4. Pinpoints where the remaining original words sit inside that area.
5. Moves the translation to the original line's level and shrinks the plate.
6. Builds a subtitle file that preserves letter joining and text direction.
7. Checks the joining and the direction in a close-up.
8. Fixes the punctuation and closes a short timing gap near the end.
9. Redesigns the subtitle in a bold face below the original line, with no plate.
10. Checks in close-up that the faces stay visible and no box remains.
11. Finds the closest clean, flat typeface to the style you asked for.
12. Re-exports with that face and reviews it across the video.
13. Pulls the frame you flagged to see why the text left the safe area.
14. Adjusts the size and the wrapping so long lines stay inside the frame.

Step one is the measurement everything else rests on. "Under the existing line" is a position relative to something that is part of the picture, not part of the subtitle track, so its coordinates have to be found on a frame rather than assumed.

Step seven is the check that only matters in some scripts. A face that renders each letter separately, or lays them out in the wrong direction, produces text that is technically present and unreadable — and at normal size it looks fine.

What does the result look like?

The translation sitting neatly below the line that was already there, in a bold flat face that reads against the picture without a plate behind it, with the faces clear and no line running off the edge of the frame.

Both languages are legible and neither is in the way.

When does this not work well?

Removing the plate means relying on contrast. Over bright, busy or pale footage a bold face is not always enough, and the honest options are a subtle shadow, a partial plate, or accepting that some shots need one.

• Bright or busy footage. Bold alone may not carry.
• A very low original line. There may be no room underneath.
• Long translations. Two lines where the original had one.
• Scripts needing special shaping. The typeface has to support them.
• Burnt-in originals you want gone. That is a removal job, not a placement one.

How do you fix a result that came back wrong?

Point at the moment where it goes wrong and say what you can see there. Vizard Agent keeps the coordinates it measured and the subtitle file it built, so a change of size, position or wrapping is a small adjustment rather than a re-translation of the whole video.

• "It runs off the edge." Size and wrapping adjusted.
• "The letters are separated." A face that shapes them properly.
• "It still overlaps the original." Re-measured and moved down.
• "I cannot read it on the bright shot." A soft shadow added.

How does Vizard Agent compare to doing it yourself?

By hand the translation is placed where subtitles usually go, which is where the original already is, and the plate is left on because removing it makes some shots unreadable. Both decisions are reasonable and both are what you are complaining about.

Common questions

Can it read where the original subtitle is? Yes. Vizard Agent measures its coordinates on a frame.

Will it cover the faces? Not without a plate. Vizard Agent checks close up.

What replaces the black box? A bold flat face, and a soft shadow if the shot demands it.

Can it match a particular typeface? Vizard Agent finds the closest available and shows you.

What about right-to-left scripts? Vizard Agent checks direction and letter joining specifically.

Can I get the subtitle file? Yes. Vizard Agent delivers it alongside the video.

Will long lines fit? Vizard Agent adjusts size and wrapping so they stay in frame.

Can it remove the original subtitle instead? That is a different job, and Vizard Agent finds it harder over moving footage.

Does the audio change? No. This is a picture-only pass.

Can it fix the timing too? Yes. Vizard Agent closed a gap near the end in this case.

What if a shot is too bright? Say so and Vizard Agent adds a shadow for those shots.

Can it do two translations at once? Yes, though Vizard Agent finds the frame gets crowded quickly.

Will the original line still be readable? Yes. Vizard Agent places under it, not across it.

Why measure rather than use the standard position? Because the standard position is exactly where the existing line already is, which is how it ended up on top of it.