LinkedIn video ad: cut a 3-second bumper from a longer video

LinkedIn's Videos API page lists video files from three seconds to 30 minutes, 75kb to 500MB, in MP4 (read 2026-10-06). To make a 3-second bumper from a finished video, call POST /v1/video-trim with start and duration: 3. The trim returns a new MP4 and leaves the source alone; one job is $0.02 at the public rate on Sume's video trim page.

Sources: LinkedIn Videos API, Microsoft Learn (read 2026-10-06) for the file limits, Video trim: cut a range from one Sume-hosted clip for Sume. That page is the Videos API reference, not an ad-format spec, and it points to LinkedIn's separate Video Ads Advertising Specifications help page; check that page before you build a bumper for a specific ad format.

What counts as a good 3 seconds?

Three seconds is one idea: a logo and a line, one product shot with a price, or the single best reaction from a longer clip. Pick the moment by looking at stills, not by guessing a timestamp. POST /v1/video-frames pulls up to 24 frames at times you name, from a source up to 300 seconds, so you can choose the moment first and cut second. See Video frames: stills at times you name.

The cut

Trim requires start and exactly one of end or duration. For exactness leave precision at its default, exact, a frame-accurate re-encode. keyframe is a stream copy and can start a GOP early, which at 3 seconds is a visible error. output can conform width, height and fps on an exact cut. The clip must already be on media.sume.com; trim does not fetch other hosts.

import json, os, urllib.request

body = {
    "video_url": "https://media.sume.com/artifacts/artf_demo/webinar-clip.mp4",
    "start": 41.2,
    "duration": 3,
    "precision": "exact",
}
req = urllib.request.Request(
    "https://api.sume.com/v1/video-trim",
    data=json.dumps(body).encode(),
    headers={"Authorization": "Bearer " + os.environ["SUME_API_KEY"],
             "Content-Type": "application/json",
             "Idempotency-Key": "linkedin-bumper-001"},
)
with urllib.request.urlopen(req) as r:
    print(json.load(r))

Check the result against the file limits

When the job reaches result_ready, GET /v1/jobs/{id}/result returns video_url, duration_seconds and actual_start_seconds. A 3-second MP4 is far under the 500MB ceiling, but 75kb is a floor too, so a very small, very low-motion clip is worth a size check. If the result carries a trim_clamped_to_source warning, your range ran past the end of the source.

A note on the surrounding changes

The same page says Marketing version 202510 sunsets on 2026-10-15, so if your upload code pins that version, move it. The stored post on the 202510 sunset covers what to change.