Vizard Agent

How to turn a horizontal video into a vertical one

Last updated 2026-08-16 · 5 min read

Upload the horizontal video to Vizard Agent and say which vertical feed it is for. Vizard Agent follows the subject through the frame instead of cropping the centre, and finds any text that was burned into the original so it still reads at the new shape. What comes back is a 1080x1920 cut you can post directly.

What is the short version?

A wide frame carries its information across its width. A vertical frame has no width to spare, so something must be chosen and something must be moved. Cropping the centre is the version that loses the speaker the moment they step sideways, and it is what most reframing tools do.

  1. Go to Vizard Agent and upload the horizontal video.
  2. Say which platform it is for, and how long you want it.
  3. Name anything that must stay readable, especially text already in the picture.

What do you need before you start?

The finished wide cut and a destination. Nothing needs pre-cropping or re-exporting, because Vizard Agent works from the video you already have. What genuinely changes the outcome is telling Vizard Agent what is in the picture besides the speaker, since text laid out for a wide frame is the thing that breaks first.

What do you type into Vizard Agent?

One sentence covers it, and the clause about text is the one people leave out. Vizard Agent will look for burned-in titles either way, but naming them means it treats them as something to preserve rather than something it found.

Prompt

Turn this horizontal video into a vertical 9:16 cut for [TikTok]. Follow the speaker, keep the title cards readable, and keep it under [60] seconds.

Variants worth knowing:

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real wide-to-vertical job. What stands out is how much of the run is about the text rather than the picture, which is the part a straight crop never addresses.

  1. Looks at the video and pulls both frames and the transcript.
  2. Reads the transcript through to the ending, so the cut lands on a finished sentence rather than a timecode.
  3. Finds exactly when the title cards appear, then zooms into them to pin the timing down to the frame.
  4. Looks at the title card design — typeface, placement, colour — because whatever replaces it has to match.
  5. Looks at the outro text separately, since end cards are laid out differently from titles.
  6. Reframes the picture, following the subject rather than fixing the crop.

Steps 3 to 5 are the reason this is not a crop. Text burned into a wide frame cannot survive a vertical crop, so Vizard Agent has to find it, understand it, and rebuild it at the new shape.

What does the result look like?

From the run this page is written from, probed on the delivered file: 1080x1920, H.264, 58.27 seconds, audio preserved. The frame rate came back at 23.976 fps, which is what the source was shot at, so the cinema rate was carried through rather than normalised to 30.

That cuts both ways and is worth saying plainly. Vizard Agent picks output specs from the source and the destination rather than from a fixed template, so if you need something particular — a frame rate, a bitrate, a duration — say it in the prompt.

This category was not separately measured, so treat the timing as a range rather than a promise: comparable work runs a median of 28 to 38 minutes end to end. Across all projects the median cost by tier is Flash 47, Pro 55, Max 242, Ultra 263 credits.

When does this not work well?

Reframing takes information away, and there are shots where the information is the width. Vizard Agent will still deliver a vertical cut of them; it will just be a worse video than the wide one, and that is worth knowing before you ask.

How do you fix a result that came back wrong?

Say what is wrong and where, in the same conversation. Vizard Agent still has the source, the transcript and the frame analysis, so a correction is a change to the existing cut rather than a fresh start, and it does not re-upload or re-transcribe anything.

How does Vizard Agent compare to doing it yourself?

Reframing by hand in Premiere or CapCut is a solved problem, in the sense that anyone can do it. What it costs is attention: a keyframed crop that follows the speaker, then every burned-in title rebuilt at the new shape, then an export to check it. Vizard Agent does the same work and reports what it changed.

By hand Vizard Agent
Reframing Keyframe the crop across every move Follows the subject automatically
Burned-in titles Rebuild each card in the new shape Finds them, matches the design, rebuilds
Finding the ending Scrub for a clean sentence Reads the transcript
Fixing one thing Re-render the export Say what is wrong, in words

The manual version is not difficult. It is repetitive, and the repetition is per title card and per movement, which is why wide videos so often stay wide.

Common questions

Will it just crop the middle? No. Vizard Agent follows the subject through the frame. Centre-crop is exactly what makes a speaker walk out of shot.

Can it go the other way, vertical to horizontal? Yes, and it is the harder direction, because the sides have to be filled rather than chosen. Tell Vizard Agent what you want at the edges: blur, a flat colour, or generated extension.

Does the audio change? No. Reframing is a picture operation and Vizard Agent leaves the audio as recorded.

Can it add captions while reframing? Yes, and doing both in one run is worth it. Captions laid out for a vertical frame are not the same as captions squeezed down from a wide one.