Change video aspect ratio by cropping or padding

Change video aspect ratio without handing the file to anyone

A shape change is a small edit with a big consequence: a 16:9 recording pushed into a 9:16 feed either loses the sides of every frame or gains two bars. Most tools online make that choice for you, quietly, after the upload finishes. This page makes it first, in words and in pixels, and then does the work in the tab you already have open. The video never leaves the device.

Pick a shape — 9:16 for Reels, Shorts and TikTok, 16:9 for YouTube, 1:1 for square feeds, 4:5 for Instagram portrait, or two numbers of your own — and then pick what happens to the picture. Those are two separate decisions and the tool refuses to merge them, because merging them is exactly how a caption ends up sliced in half without anyone noticing until it is published.

No shape is selected for you. The moment you choose one, the page shows the exact output size and a preview of a frame from your video in that shape, drawn with the framing you have chosen. Switch the shape, the framing or the padding color and the preview redraws on the spot; the file does not have to be chosen again. The preview is drawn by the same code that later draws every frame of the export, on a smaller canvas, so it is the export in miniature rather than an approximation of it.

For Shorts the shape is not only a matter of framing: it is what decides whether YouTube files the upload as a Short at all. YouTube publishes a rule rather than a number there — square or vertical, three minutes, 1080p — and never the 1080 × 1920 every guide repeats. YouTube Shorts size quotes the four Help pages that say so, and the one Google page where 9:16 actually appears.

Cropping, a color and a blurred background are different promises

Crop to fill scales the picture until it covers the whole frame and cuts away what hangs over the edge, centered. Nothing is added and the frame is completely used, but pixels are gone for good. The page says how many before you run it: a 320 × 180 clip taken to 1:1 loses 140 px of width, which is 43.8% of the picture.

Pad with color scales the picture until it fits entirely inside the frame and fills the remainder with two bars in a color you pick. Black is only where the picker starts. Every recorded pixel survives, but part of the frame is now not your video. The same clip padded to 1:1 gains two bars of 70 px each, 43.8% of the finished frame.

Pad with blurred video keeps exactly the same picture in the middle, but fills the bars with an enlarged copy of the video itself, scaled to cover the whole frame and heavily blurred. It is redrawn for every frame, so the background moves with the footage instead of sitting still behind it. This is the look vertical feeds use for footage shot sideways: the bars stop reading as an empty border and start reading as part of the shot.

The padding is the part a tool can be sloppy about, so it is read back rather than assumed. After the export, the first frame of the finished MP4 is decoded again and the middle of each bar is sampled. For a color, the value found in the file is printed next to the one you picked; H.264 stores color at reduced precision, so a few levels of drift are normal, and anything larger is reported as a mismatch rather than hidden. For the blurred background, pixels along the bars are compared to make sure they hold a picture and not a flat fill.

The blur itself is not made with the canvas blur filter. In WebKit, the engine inside Safari, that filter was measured to do nothing at all — no error, and a razor-sharp edge where a soft one was asked for. The background is instead halved again and again until it is tiny, then enlarged back step by step, which every browser draws the same way. At very low bitrates H.264 ringing can tint the single row of pixels touching the picture edge; that is a compression artifact at a hard boundary, not a second fill.

The frame you are promised is the frame that gets measured

Both output edges are always even numbers. H.264 codes in pairs of pixels, so an odd edge is rounded by the encoder whether or not the interface admits it — a page that prints 101 px and writes 102 px is wrong in a way no error message will ever surface. The frame shown here is already even, so the encoder has nothing left to round.

When the export finishes, the finished MP4 is opened again and its picture size is read out of the file. The receipt shows that measured frame next to the one you asked for. If they ever disagree the receipt says so instead of repeating the request back to you, which is the only version of this claim that can be falsified.

The default frame never enlarges anything. For a crop it is the largest frame of your chosen shape that fits inside the source; for padding it is the smallest that contains it. Upscaling invents detail that was never recorded, and doing it by default would make every export look softer for no reason. When a platform insists on a specific size, type the frame width yourself and the height follows the shape.

The MP4 this tool writes

MP4

Full name	MPEG-4 Part 14
Extension	.mp4, .m4v
Format type	Video container — usually H.264 or H.265 with AAC
MIME type	video/mp4

What the re-encode changes, and what it leaves alone

Reshaping a frame means drawing it again, so the picture is decoded and re-encoded as H.264. Sound is left alone where possible: an AAC soundtrack is copied through unchanged, and anything else becomes AAC so the MP4 stays widely playable. Container metadata, chapters, subtitles and additional tracks are not carried over. If the file also needs to hit a size limit, run the local video compressor afterwards — it targets a measured byte ceiling rather than a shape, and stacking two lossy encodes is worth doing in that order.

A shape change is often not the only edit. Cutting the clip down first with the video trimmer means less footage to re-encode and a faster export. If the destination wants a looping animation rather than a video, video to GIF is the honest route, and the frame extractor pulls a single still out at full resolution when a thumbnail is what is actually needed. A WebM that only needs a different container can go through WebM to MP4 instead of being reshaped.

Drawing your own rectangle over the picture is a genuinely different job from choosing a shape, and it belongs on its own page rather than behind a toggle here. This tool answers “make it 9:16”; it does not pretend to answer “keep exactly this part of the frame”. When a face, a caption or a logo sits off-center and a centered crop would cut it, crop video is the page that lets you place the rectangle yourself and reports what it costs in pixels.

Local work still costs local resources

The source is range-read on this device, the new MP4 is streamed into origin-private browser storage when the browser permits it, and the download link points at a file that was created here. There is no URL import, no queue and no invisible server fallback; a browser missing the required decoder or the H.264 encoder is told exactly which capability is absent. You can check that boundary yourself with the network verification guide.

Long or high-resolution clips take real time and real memory, and a tab that runs out of memory reloads itself without explaining why. The size notice appears before any work starts, and no file is ever refused for being large — the machine is yours to spend. Canceling stops the running conversion and removes the partial scratch file; abandoned scratch files older than a day are cleared on a later visit.