Ad audio spec table: YouTube 48 kHz, Meta AAC 128 kbps, LinkedIn AAC

YouTube recommends AAC-LC, Opus or Eclipsa audio at 48 kHz, Meta asks for stereo AAC at 128 kbps or more, and LinkedIn lists AAC or MPEG4 audio. If you keep the audio when you trim with Sume's exact precision, it is remuxed as AAC, which is on all three lists. Sume does not let you set the audio sample rate or bitrate on trim or timeline.

What each page says

The values are copied from each page on 2026-10-08. LinkedIn's page has an audio size line that reads in kilohertz; the text is quoted as written and not interpreted.

What Sume controls

Three tools touch audio. Trim has an audio setting, keep or drop, and exact precision remuxes kept audio as AAC. Audio detach pulls the track out as a wav or mp3 at a sample rate of 16000, 44100 or 48000 Hz for $0.01 per job. Timeline sets level with audio.gain_db from -60 to 12 and can lay a soundtrack under a spine with duck_db from 0 to 20.

Picking the 48 kHz case

If your voice-over is a separate file and YouTube's 48 kHz matters to you, detach it at 48000 and use that wav as the spine. The detach job has a 1800 second source limit and a 900 second output limit, so it suits ads and Shorts, not feature films.

The sketch below detaches audio from a Sume-hosted video as wav at 48 kHz. Field names follow the audio detach page; the artifact URL is a placeholder.

curl -X POST https://api.sume.com/v1/audio-detach \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: detach-48k-001" \
  -d '{"video_url":"https://media.sume.com/artifacts/artf_demo/ad.mp4","format":"wav","sample_rate":48000}'

What is left to you

Stereo versus mono, bitrate and the choice between AAC-LC and Opus are decided by the encoder, and Sume does not expose them. If a platform review rejects an ad for audio, re-export the audio from your editor and bring it back as the spine. Sume's job is to place and cut the sound you supply, with the gain and ducking controls above. Check the result by detaching the audio from the rendered file and listening to the first and last seconds, where fades and cuts usually show a problem. The detach job is $0.01, so a check costs less than the render it verifies. Keep gain changes small and re-render rather than stacking several edits, because each pass is another encode.