LinkedIn video ad 5% aspect tolerance: crop a tall phone recording

A tall phone recording such as 1080x2160 is not a 9:16 file, and LinkedIn's video ad spec lists a 5% aspect-ratio tolerance, so plan to crop it. The width-to-height ratio of 1080x2160 is 0.5, while 9:16 is 0.5625, so the file is about 11% off before you upload. Sume's video filter can crop it to 1080x1920 for $0.02 per job.

One caveat first: the LinkedIn page lists the tolerance as 5% but does not say how that percentage is measured (read 2026-10-10). Treat any file that is not within a pixel or two of the target ratio as a risk, and crop rather than hope.

What the LinkedIn page says

The fetched help page lists four ratios, each with a size range. Width and height must each be between 360 and 1920 pixels, and the file must be MP4.

A useful habit is to compute the ratio of every file before it enters the ad set. Divide width by height and compare with 0.5625 for 9:16, 0.8 for 4:5, 1.7778 for 16:9 and 1 for square. A file that lands within a few thousandths of the target is safe by any reading of the tolerance. A file that is off by several percent is where the unstated measurement rule starts to matter, and you cannot know the answer from the page.

Why a phone recording misses

Many phones record tall screens at ratios taller than 9:16. A 1080x2160 capture has a ratio of 0.5. The 9:16 target is 0.5625. Dividing gives 0.889, so the capture is about 11% narrower than the target for its height. That is more than double the 5% on the page. The height is also 2160, above the 1920 maximum for that edge, so the file fails on two counts.

The crop that fixes it

Keep the full width and trim height from 2160 to 1920. As a fraction of the frame, that is 1920 divided by 2160, which is about 0.8889. To center the cut, the top offset is half the remainder: (1 - 0.8889) / 2, about 0.0556.

Sume's video filter takes crop as fractions: x and y in 0 to 1, width and height in 0.05 to 1, and the compiler rounds to even values for yuv420p. So a crop of x 0, y 0.0556, width 1, height 0.8889 gives a 1080x1920 frame. The source must be 300 seconds or shorter and a Sume-hosted clip, so import it with a media import first.

• Run POST /v1/video-filter/check first. It is free and returns diagnostics instead of a 400.
• Submit POST /v1/video-filter with an Idempotency-Key. The result is a new MP4 and the source stays unchanged.
• Measure the output with video inspect before you upload.

curl -X POST https://api.sume.com/v1/video-filter \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: crop-tall-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/phone.mp4",
    "ops": [{ "op": "crop", "x": 0, "y": 0.0556, "width": 1, "height": 0.8889 }]
  }'

What to check after the crop

Cropping removes about 11% of the height, 240 pixels in this example. Look at where the subject and any burned-in text sit. If the top or bottom strip holds a status bar or a caption, shift y toward that edge instead of centering. Video inspect returns the probe, so you can confirm the width, the height and the duration before you spend media budget.

The video filter keeps the frame rate and audio of the source unless the program changes them. The LinkedIn pages also disagree on frame rate: the help page says less than 30 FPS and the business page recommends 30 (read 2026-10-10), so check which one your account manager expects. Sume does not change the frame rate in a crop.

A second case is a 4:5 target. A 1080x1920 clip cropped to 4:5 at full width needs a height of 1350, a fraction of 1350 divided by 1920, about 0.7031, and the 1080x1350 size is the maximum on the page for that ratio. The same crop tool covers both, and only the height fraction changes. For 1:1, a 1080x1920 source needs a height fraction of 1080 divided by 1920, which is 0.5625.

Run the free check endpoint for each variant before you pay for the encode, because it flags a crop that leaves the frame or goes under 0.05 on a side.

Where this stops

Sume does not submit to LinkedIn and cannot tell you whether LinkedIn will accept a given file. It gets the geometry to a documented target. The tolerance math above is arithmetic on the numbers in the spec, not a statement about how LinkedIn measures it.