Pinterest video Pins: crop a 9:16 master to 1:2, 2:3, 3:4, 4:5

Pinterest recommends 9:16 for full-bleed video Pins and lists 1:2, 2:3, 3:4, 4:5 and 1:1 as other options. From a 9:16 master, a centred crop to 3:4 keeps 0.75 of the height, and 2:3 keeps 0.84375; Sume's video-filter crop takes these as fractions.

The ratio list is from Pinterest's Pin specs page (read 2026-10-10). The fractions are my arithmetic using the Video filter docs.

What Pinterest lists

For full-bleed videos the page recommends a 9:16 ratio or 1080 x 1920 pixels, with the other options 1:2, 2:3, 3:4, 4:5 or 1:1. It lists a length of minimum 4 seconds and maximum 5 minutes, file types .MP4 and .M4V on the web and .MP4, .MOV and .M4V on Android and iOS, and H.264 or H.265 video encoding. The page gives no maximum file size for video Pins in the version I fetched.

The crop fractions

A 9:16 frame is 0.5625 wide per unit of height. A target taller than 9:16 is not possible by cropping, and a target wider is reached by cutting height. For target ratio r wider than 0.5625, the kept height fraction is 0.5625 divided by r, centred at (1 minus that) / 2. For 1:2, which is narrower than 9:16 at 0.5, the width is cut instead: 0.5 / 0.5625 = 0.8889, starting at 0.0556.

Sending it to Sume

The Video filter docs say crop uses fractions of the source frame, with width and height in 0.05 to 1, and that the compiler rounds to even values. On a 1080 x 1920 source the 3:4 crop is 1080 x 1440 and the 4:5 crop is 1080 x 1350, both exact and even. 2:3 gives 1080 x 1620. 1:2 gives 960 x 1920. The check endpoint is free, and the encode is $0.02 per job with a source limit of 300 seconds.

Re-compose rather than just crop if the subject sits off-centre. The y offset in the table is centred; shift it to keep a face inside the frame, as long as y plus height stays at or under 1.

• Check with POST /v1/video-filter/check.
• Encode with an Idempotency-Key.
• Use the new video_url, never the source.

Length and codec

The 4 second minimum and 5 minute maximum are separate from the crop. A Sume trim can bring a longer source down; its output goes from 0.2 to 900 seconds, so use duration 300 or less for Pinterest. The page lists H.264 or H.265. Sume's exact trim re-encodes with libx264 and yuv420p, which is H.264 per the Video trim docs; Timeline's codec is not stated in the page I read, so check a rendered file if the codec matters to you.

Not every crop is a good idea. Cutting a 9:16 clip to 1:1 removes more than 40 percent of the height, which can chop off a caption area or a product at the bottom. A centred 1:1 crop trims about 22 percent from the top and 22 percent from the bottom, so check where your text and product sit before you commit.

So compare two options: crop the vertical master, or render a purpose-built version through Timeline from the original clips with cover fit at the target size. The second costs a render by the output minute, rounded up, but it lets Timeline fill the frame from the source rather than from an already-cropped file.

For a batch, produce the nine by sixteen master once and derive the others with the free check call first. Each derived crop is a $0.02 encode, so five variants cost ten cents, and you can compare them side by side before deciding which ratio to publish on Pinterest.