LinkedIn video ad specs: under 30 fps, 4:5, and SRT captions

LinkedIn video ads take an MP4 in H.264 or VP8 at less than 30 fps, AAC or MPEG4 audio, 3 seconds to 30 minutes, and 75 KB to 500 MB, with 4:5 as the recommended ratio (at most 1080×1350) and captions only as an SRT file. With Sume, set output.fps to 24 or 25, crop 9:16 clips to 4:5 with video filter, and write the SRT yourself from an inspect transcript, because Sume burns captions into the picture.

LinkedIn's rules are quoted from its Video ads advertising specifications help page and the Videos API docs on Microsoft Learn, read on 2026-09-27. Sume facts come from Timeline 1.0, Video trim, Video filter, and Video inspect.

What are LinkedIn's video ad specs?

LinkedIn's two pages set these file rules, and the table below gives each ratio's pixel range, with 4:5 marked as the recommended ratio:

• Length: three seconds to 30 minutes. LinkedIn recommends 15–30 seconds so an ad can qualify for every placement across the LinkedIn Feed and the Audience Network.
• File: MP4, H.264 or VP8, frame rate “Less than 30 FPS”, audio AAC or MPEG4, and an audio row that reads “Less than 64KHz”.
• Size: 75 KB to 500 MB on both pages, though the Videos API also says “Maximum allowed Videos size is 5GB”. Plan for 500 MB.
• Captions: SRT only, text only, no custom formatting. The Videos API takes one caption file per video, in English only.

How do I get a Sume video under 30 fps?

Set the rate explicitly. Timeline and trim take output.fps of 24, 25, 30, or 60, and only 24 and 25 are below 30. Don't omit it: an omitted Timeline rate follows the sources, and generated clips don't document theirs. In Timeline, 24 or 25 over faster sources repeats or drops frames, which judders on motion (how Timeline picks a rate). Probe the result for free with a frames: false inspect to confirm fps.

For one clip, trim's output can carry just fps:

curl -X POST https://api.sume.com/v1/video-trim \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: linkedin-25fps-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/clip.mp4",
    "start": 0,
    "duration": 20,
    "output": { "fps": 25 }
  }'

How do I make the recommended 4:5 version?

No Sume video model takes a 4:5 aspect_ratio, so derive it from 9:16. Send one video filter job with a crop op of x: 0, y: 0.1484375, width: 1, height: 0.703125 (the 4:5 arithmetic is in the Meta ad specs post) and a filtergraph of scale=1080:1350,fps=25. Both filters are on the allowlist, the audio carries over, and sources can run up to 300 seconds.

Longer or multi-clip ads go through Timeline with output: { width: 1080, height: 1350, fps: 25 }, LinkedIn's 4:5 maximum. Today a render's sound comes from its audio spine, so detach the clip audio first. The 16:9 and 9:16 maxima match Timeline sizes too: 1920×1080 and the default 1080×1920.

How do I add LinkedIn captions to a Sume video?

Sume's caption API burns text into the picture, and its docs list a captioned video_url, not a caption file, so build the SRT yourself. Run POST /v1/video-inspect with transcribe: true and segmentation: { mode: "sentence" }. The transcript's segments[] carry index, text, start, and end and are shaped as caption lines; write each as one SRT cue. Only the transcript is billed, per audio minute (transcript steps).

For Korean viewers, the Videos API's English-only captions don't help. Burn Hangul captions into the video with a Hangul caption style instead (Korean captions); today one caption job takes a source of at most 60 seconds.

How does the file get to LinkedIn?

LinkedIn says video ads must contain a video file uploaded directly to LinkedIn, not a link. The Videos API's initializeUpload takes the file size and returns upload URLs for 4 MB byte ranges; you PUT each part, keep the ETags, and call finalizeUpload. So download the MP4 from its media.sume.com URL and upload the bytes. A thumbnail is optional, JPG or PNG up to 2 MB, and should match the video's ratio and resolution; POST /v1/video-frames returns JPEG or PNG stills at the source size from clips up to 300 seconds.