TikTok reservation ads must include audio: probe, then add a bed

A silent clip is a failed reservation creative. TikTok's reservation in-feed page says all video creatives must include audio (read 2026-10-07). Before upload, probe the file with Sume's video-inspect and read has_audio; if it is false, assemble the clip again with a voice or music spine in Timeline 1.0.

The check is cheap and the fix is one render. The part to avoid is declaring silence: a Timeline audio.mode: "silence" produces a declared-length track with no file, which is the opposite of what the page asks for.

What the page says

The same reservation page lists 5-60 seconds with 9-15 recommended, a 500 MB cap, and a 2,500 kbps floor. It also says watermarks, including TikTok's own branding, are prohibited. The audio line is short; it does not name a loudness target, a codec, or a minimum track length, so I do not invent one.

Step 1: probe with frames off

Send the clip URL with frames: false. The clip must already be a media.sume.com artifact or asset of your workspace; import it first with POST /v1/media-imports, per Media inputs. Read probe.has_audio in the answer.

curl -X POST https://api.sume.com/v1/video-inspect \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: audio-probe-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/clip.mp4",
    "frames": false
  }'

Step 2: give the clip a spine

Timeline 1.0 takes one audio spine plus ordered video[] slots. For a 12-second ad, send audio.url (a voiceover or a track you made, hosted on Sume) with audio.duration_seconds: 12, and place the clip in video[0] with start: 0. Optionally add a soundtrack bed with duck_db so the music dips under speech; ducking needs a real spine, not silence.

Music 1.0 makes a track from a text prompt. It does not take a duration field, so ask for the length in the prompt and trim in the timeline. The public Timeline rate is $0.10 per ceil output minute, so a 12-second ad reserves one minute (confirm in GET /v1/catalog).

Limits

Sume does not inspect your account or submit anything to TikTok. A passing has_audio only proves a track exists; whether the sound is acceptable to reviewers is TikTok's call. Also watch the audio you inherit from a downloaded clip: the watermark rule above applies to the pictures you reuse.

A quick decision list

If has_audio is true and the track is the voice you want, upload. If it is true but the audio came from a stock clip you cannot clear, replace it. If it is false, build the spine first, then the picture.

Mind sync when you add a spine to existing footage: the timeline places slots on the spine's clock, and video[0].start must be 0, so cut the picture to the voice rather than the other way round. A 12-second voice track means 12 seconds of slots, with coverage allowed to stop at most 0.5 seconds early.