Cut away to B-roll without breaking the voice: one detached spine

A B-roll cutaway keeps the voice running underneath while the picture changes. In a Timeline render the audio spine is independent of the video slots, so you detach the A-roll audio once ($0.01), list three video slots (A, B, A again) and let the spine play straight through; a 32-second Reel renders for $0.10, $0.11 in all.

The document

The A-roll slot returns at source_in 19.5 and start 19.5, so the speaker's lips stay in step with the spine because source time and output time advance together. The B-roll slot covers 14 to 19.5 s and its own sound is ignored, since the spine carries the voice. Slots must be listed with start increasing from 0.

{
  "audio": { "url": "https://media.sume.com/artifacts/artf_demo/a-roll.wav", "duration_seconds": 32 },
  "video": [
    { "source_url": "https://media.sume.com/artifacts/artf_demo/a-roll.mp4", "source_in": 0,    "start": 0,    "duration": 14 },
    { "source_url": "https://media.sume.com/artifacts/artf_demo/b-roll.mp4", "source_in": 0,    "start": 14,   "duration": 5.5 },
    { "source_url": "https://media.sume.com/artifacts/artf_demo/a-roll.mp4", "source_in": 19.5, "start": 19.5, "duration": 12.5 }
  ]
}

Cost

Prices are the flat per-job rates from the docs; the render bills per ceil output minute, and the 32-second result is one minute.

What to check

• Use the detached wav as the spine. Do not use the 16 kHz mono transcript.audio_url from video inspect; the render flags it with audio_spine_low_fidelity.
• Make audio.duration_seconds at most the length of the audio, here 32 s.
• If B-roll is shorter than its slot the render pads or loops it and warns; run POST /v1/timeline-1.0/plan first for the free duration and segment check, but remember the plan does not download media.
• The B-roll must be a media.sume.com artifact; import other footage first.