How to join clips whose frames are different sizes without stretching them

Last updated 2026-09-17 · 8 min read

What is the short version?

Clips from different phones, different apps and different crops tend to end up the same width and three different heights. Forced onto a single canvas they either stretch, which distorts every face and every line of text in them, or they sit inside black bars, which looks unfinished rather than deliberate.

1. Ask Vizard Agent to measure all the clips before anything is joined.
2. Say nothing may be scaled to fit.
3. Ask for the leftover space filled from the picture itself.

What do you need before you start?

Every clip, and the shape you want out at the end. Vizard Agent chooses the canvas from what it actually measures across the whole set, so having all of the clips in hand before it starts is what avoids ending up with a canvas that suits three of the four.

• All the clips. Not a first batch.
• The output shape. Usually 9:16 or 16:9.
• The order. They are being joined, after all.
• What must not distort. Faces, text, logos.
• Whether transitions are wanted. They complicate unequal lengths.

What do you type into Vizard Agent?

Rule out stretching in a sentence of its own. Squeezing everything to one frame is what Vizard Agent would otherwise default to, because it is the only way to make four differently shaped clips identical — and it is the single outcome you cannot accept.

Prompt

Variants worth knowing:

• "Nothing may be deformed." The constraint, stated flatly.
• "Not the speed." Frame rates differ too.
• "A soft background taken from that clip." Better than black.

What does Vizard Agent actually do?

Here is the Vizard Agent sequence on four vertical clips that shared a width and had three different heights between them. The first version is honest and slightly ugly — it has empty fields in it — and the fix for that is the interesting step.

1. Checks the parameters of all four clips.
2. Surveys their frames to choose a tidy layout.
3. Confirms every source file is actually reachable and complete.
4. Reviews the four sources together as one contact sheet.
5. Checks the audio tracks and the intended order.
6. Joins them into one video with nothing deformed.
7. Inspects the start, the joins and the ending.
8. Replaces the empty fields with a soft background from the frames themselves.
9. Normalises the frame parameters and renders again.
10. Adds overlapping transitions and checks them before, during and after.

Step eight is the one that turns a compromise into a design. A clip that is shorter than the canvas leaves space, and space filled with a blurred, darkened version of that same clip reads as an intentional treatment rather than as something missing.

Step one is not a formality. A container can report a size the picture does not actually have, so the layout has to be built on measured dimensions rather than on what each file claims about itself.

Step ten is harder than it looks with unequal clips. An overlapping transition borrows time from both sides, and clips of different lengths do not all have that time to give.

What does the result look like?

One video where every clip keeps its proportions. Faces are the right shape, text is not squashed, and the frames that do not fill the canvas sit in a soft field of their own colours rather than in black bars.

When does this not work well?

Very different shapes cannot be reconciled kindly. Mixing a tall vertical with a wide landscape on one canvas leaves so much empty space that the background treatment becomes most of the screen, and cutting rather than fitting is the better answer.

• Mixed orientations. Too much left over to fill.
• Very low-resolution clips. They look soft when fitted.
• Clips with their own bars. Those need trimming first.
• Text near the edges. The fit can push it off the canvas.
• Wildly different frame rates. They need normalising first.

How do you fix a result that came back wrong?

Say which clip it is and what you can see. Vizard Agent measured each one separately and fitted each one separately, so a note about the third clip changes only how that clip sits on the canvas and leaves the other three exactly as they are.

• "The third one looks squashed." Its real dimensions are re-measured.
• "There are black bars." The background fill is applied to that clip.
• "The background is distracting." It is darkened or softened further.
• "The transition stutters." The overlap is shortened to what the shorter clip allows.

How does Vizard Agent compare to doing it yourself?

By hand you set a sequence size, drop the clips in, and the editor scales them to fit because that is what editors do. It looks nearly right, and three months later somebody points out that everyone in clip two is slightly wide.

Common questions

Why not just scale them to fit? Because that deforms faces and text. Vizard Agent fits them honestly instead.

What canvas will it pick? One that suits the measured clips. Vizard Agent tells you what it chose and why.

Is a blurred background better than black? Usually, and it reads as intentional. Ask Vizard Agent for either.

What if the clips are different frame rates? Vizard Agent normalises them rather than letting the speeds drift.

Can it crop instead of fitting? Yes, if you prefer losing edges to filling space. Say which.

Will the audio line up? Yes. Vizard Agent checks the tracks before joining.

What about clips that already have bars? Those are trimmed first, or you get bars inside bars.

Can it add transitions? Yes, though unequal lengths limit how long they can be.

How many clips can it join? As many as you have. Vizard Agent measures each one.

What if one clip is much lower quality? Vizard Agent flags it; fitting makes soft footage look softer.

Will text near the edge survive? Check it. Vizard Agent inspects the edges after fitting.

Can I see the measurements? Yes. Ask Vizard Agent for the real dimensions of each source.

Does the order matter? Only editorially. Tell Vizard Agent the sequence you want.

What if I add another clip later? Vizard Agent re-measures and can keep the same canvas if it still suits.