16:9 to a 9:16 Short: crop first with Sume, then trim to 1080x1920

To make a vertical Short from a 16:9 video with Sume, crop first and resize second. Use video filter with a center crop, then video trim with output set to 1080x1920. Skipping the crop would stretch the whole wide frame into a tall one, because the trim conform is only a scale plus frame rate.

What YouTube says about Shorts

YouTube's Shorts help page gives the limits for the upload side.

Step 1: the center crop

For a 1920x1080 source, a 9:16 window is about 608 pixels wide and full height. As fractions of the frame, that is width: 0.3164, height: 1, y: 0 and x: 0.3418 to center it. Video filter's crop op takes exactly these fractions. Send the crop in ops[], run the free /v1/video-filter/check first, then submit the job.

If the speaker is not in the middle, move x; the only rule is that x + width stays at or under 1.

{
  "video_url": "https://media.sume.com/artifacts/artf_demo/talk.mp4",
  "ops": [
    { "op": "crop", "x": 0.3418, "y": 0, "width": 0.3164, "height": 1 }
  ]
}

Step 2: resize and cut with video trim

The filter result is a new media.sume.com artifact, so pass it as the video_url to POST /v1/video-trim with start, duration and output: { width: 1080, height: 1920, fps: 30 }. The crop is already 9:16, so the scale does not distort it. Output sizes are 256 to 2160, so 1080x1920 is in range, and output needs exact precision, the default.

Be honest about quality: a 608x1080 crop scaled to 1080x1920 is an upscale, so a source at 1080p will look softer than native vertical footage. A 4K source holds up better. Keep the clip under three minutes.

Limits

This is a fixed center or chosen-position crop, not a face-tracking reframe. If the subject moves across the frame, cut the clip into parts and crop each with its own x. Each job is billed separately, at $0.02 each in the public docs; confirm in GET /v1/catalog.