How to make your subtitles smaller without making them unreadable
Ask Vizard Agent for smaller subtitles and say they still have to be readable — those are two instructions, not one. It picks a typeface that holds its shape at the smaller size, keeps the lines short so they do not wrap, and checks the result on rendered frames instead of a preview.
What is the short version?
Caption presets are sized for a phone held at arm's length with the sound off, which is the right call for a talking head and the wrong one for landscape footage where the picture is the point. Vizard Agent can set them smaller, but only if you also say how small is too small.
- Tell Vizard Agent to set the subtitles smaller but keep them readable.
- Say what they must not cover — the horizon, a sign, a face.
- Ask to see rendered frames rather than trusting a preview.
What do you need before you start?
The video and a sense of where the text is in the way. Vizard Agent can measure legibility, but "they cover the view in the drone shots" is the kind of instruction that changes the position as well as the size.
- The video. With or without captions already on it.
- How much smaller. A little, or as small as still works.
- What must stay visible. The scenery, a sign, a logo.
- The language. It decides the typeface.
- Any spellings to fix. Place names, brand names.
What do you type into Vizard Agent?
Put the constraint in the same sentence as the request. "Smaller" on its own invites a size that technically renders and cannot be read on a phone, and the whole point of the change is that people still read them.
Prompt
Variants worth knowing:
- "Still readable on a phone." The limit on how small.
- "Not over the scenery." Position, not just size.
- "Same length." Stops the edit being rebuilt around the change.
What does Vizard Agent actually do?
Here is the Vizard Agent sequence on a three-minute travel video whose captions were too heavy for the footage. The size is the easy part; the typeface and the line lengths are what make the smaller size work.
- Watches the video and pulls frames to see the caption layout and the ending.
- Transcribes the speech so the new subtitles match what is actually said.
- Checks the opening, middle and closing scenes for the place name spelling.
- Enlarges the title frames to read the brand name exactly.
- Chooses a typeface that stays legible small in that language, with its diacritics intact.
- Builds a subtitle file with the corrected spelling and a compact style.
- Checks how the tool laid out the lines, because a small size wraps differently.
- Renders, then inspects frames from the start, middle and end.
- Covers the old caption layer properly where the new smaller text no longer hides it.
- Rebuilds the ending on a high-resolution drone shot with the corrected sign-off line.
- Exports at exactly the original length and re-checks the frames.
Step five is the decision that makes this work. Below a certain size a typeface with thin strokes or tight counters turns to mush, and in a language with stacked diacritics the marks are the first thing to disappear — so the size and the face are chosen together.
Step nine is the trap in this job specifically. If there were captions burnt into the video already, smaller replacements do not cover them, and you end up with the old line peeking out from behind the new one.
Step seven is worth asking for. A smaller size fits more characters per line, which means the tool re-wraps everything, and a line that was two lines before may now break in an awkward place.
What does the result look like?
Subtitles that sit lower and lighter in the frame, that you can still read at arm's length, with the diacritics and the corrected spellings intact. The scenery is visible again, the old caption layer is gone rather than hidden, and the video is the same length it was.
When does this not work well?
Some footage cannot take small captions at all. Over busy, bright or high-contrast pictures, small text needs an outline or a shadow to survive, and past a certain point the outline is doing the work that the size was supposed to remove.
- Bright or busy backgrounds. Small text needs help to be seen.
- Long sentences. Smaller does not mean more words per line.
- Languages with stacked marks. The diacritics go first.
- Very low resolution. There are not enough pixels for small text.
- Viewers watching on a watch or a feed preview. Smaller is worse there.
How do you fix a result that came back wrong?
Say whether the problem is reading the subtitles or seeing past them. Those two complaints pull in opposite directions, so Vizard Agent adjusts the size, the weight or the position depending on which one you name — and guessing between them costs a render.
- "Too small to read now." The size goes back up a step, with more weight.
- "They still cover the view." They are moved rather than shrunk further.
- "The old captions show behind." The covering layer is rebuilt.
- "The accents have vanished." A typeface with stronger diacritics is chosen.
How does Vizard Agent compare to doing it yourself?
By hand you drag the size slider down and watch it in a preview window that is half the size of a phone. It looks fine. Then you watch it on a phone and half the lines are unreadable, and the ones you can read have lost their accents.
| By hand | Vizard Agent | |
|---|---|---|
| Size | A slider, judged in a preview | Set with the typeface, checked on frames |
| Typeface | Whatever the preset uses | Chosen for legibility at that size |
| Line breaks | Re-wrapped silently | Inspected after the change |
| Old captions | Left showing behind | Covered deliberately |
| Spellings | Unchanged | Read off the frames and corrected |
Common questions
How small can subtitles go? Until they stop being readable on a phone — Vizard Agent checks that rather than guessing.
Can they be moved instead? Often moving is the better fix, and Vizard Agent will suggest it when it is.
Will the accents survive? Yes, if the typeface is chosen for it — Vizard Agent treats that as part of the job.
Can I keep the same words? Yes. Vizard Agent changes only the presentation unless you ask otherwise.
What if captions are already burnt in? Vizard Agent covers the old layer rather than leaving it visible behind the new text.
Can it fix spellings at the same time? Yes, and it reads them off the frames to confirm the correct form.
Does the video stay the same length? Yes, if you say so — Vizard Agent exports to the original duration.
Can it use my brand font? Yes. Supply the file and Vizard Agent tests it at the smaller size.
What about an outline or a shadow? Vizard Agent can add either, and small text over bright footage usually needs one.
Can the ending be redesigned too? Yes. Vizard Agent can build a three-second closing card from your own footage.
Will it work for two languages at once? Yes, though two lines at a small size is the hardest case to keep readable.
How do I check it myself? Watch it on a phone, not in the editor — that is the only test that counts.
Can it match the size of another video? Yes. Give Vizard Agent the reference and it measures the text in it.
Does smaller text help the algorithm? No. It helps the viewer see the footage, which is a different and better reason.