Audience Network in-stream: 5-30 seconds, sound required, keep audio

Short answer

In Meta's video requirements chart, Audience Network in-stream is the placement that lists a length of 5 to 30 seconds and the one where sound is required. Meta lists sound as optional for the Stories and Instagram placements and as optional but recommended for Facebook Feed and Marketplace. If you export one silent clip for all placements, the in-stream version needs audio added.

With Sume Video Trim, audio defaults to keep, so a trim of a clip that has sound stays audible. The risk is a clip that never had sound, such as many generated videos, or a trim where someone set audio to drop for a silent autoplay version.

What the chart says

Only a few rows matter here. The table compares in-stream with the placements most advertisers run together.

Making the in-stream version

Treat the in-stream cut as its own deliverable instead of a trim of the silent one. Build it in Timeline with a soundtrack or a voiceover: the soundtrack object takes a url, gain_db, loop, fade_out_seconds (up to 10) and duck_db (0 to 20), so music can sit under a voice. Then check duration is between 5 and 30 seconds before you submit it.

If you start from a clip that already has sound, trim it with Video Trim and leave audio at keep. In exact mode the kept audio is remuxed as AAC. If the clip has no audio at all, trimming will not invent any, so add the track in Timeline first.

• Run POST /v1/timeline-1.0/plan to confirm duration_seconds lands in 5-30.
• Use audio keep when trimming from a sound-bearing master.
• Do not reuse the audio-dropped autoplay version for in-stream.
• Add a fade-out so a hard cut at 30 seconds does not sound abrupt (output.fade_out_seconds is 0 to 5).

Checks before upload

Verify the final file has an audio stream, not just that you did not choose drop. A quick ffprobe listing the streams is enough. Check duration against 5 to 30 seconds, and give yourself a margin: a 30.0-second target that renders as 30.4 seconds is out of range.

The chart is Meta's own summary, so recheck it when you start a campaign; the placement list changes over time. For the wider length table see Meta video length by placement.

Where the silent version still belongs

Keep the silent, captioned version for placements that autoplay muted, such as feeds. Many viewers see video with sound off, so a version with burned-in captions is useful there. Sume's Video Captions job can burn captions onto an existing video URL in styles like slam, punch, black-outline and tiktok-green, so the same footage can serve both: captions for the muted feed, music or voice for in-stream.

Keep them as separate files with clear names, such as the placement and length in the filename. A mix-up between the silent cut and the in-stream cut is an easy mistake in a folder of near-identical clips, and it is expensive if it only shows up after the ad is in review.