Dental clinic explainer: a calm AI music bed and its cost

For a clinic or dental explainer, ask for a calm, unhurried instrumental that never draws attention, about 90 seconds long, and duck it under the voiceover. On Sume that is one Music Router generation ($0.125) plus a two-minute Timeline render ($0.20, because 90 seconds bills as two started minutes), so $0.325 for the audio and assembly.

Healthcare explainers have a particular job for music. Viewers may be nervous, so the bed should feel steady and light rather than clinical or dramatic. It also sits under a spoken script that carries the facts, so the track must stay out of the speech band as far as a prompt can ask for that.

The brief to send

The brief avoids vocals and spoken word, sets a slow tempo, and keeps a single gentle moment at the midpoint where the script moves from the problem to the solution.

curl -X POST https://api.sume.com/v1/music-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: dental-clinic-explainer-video-music-calm-001" \
  -d '{
      "model": "sume/music-auto",
      "prompt": "Calm reassuring background for a clinic explainer, 72 BPM, G major. Soft piano, warm pad, light acoustic guitar harmonics, very little percussion. A slight brightening at 0:45 as the explanation turns hopeful, then a steady, quiet finish by 1:30. Clean, modern, close production. No spoken word. Instrumental, no vocals. A 90-second track."
    }'

Why each axis is set that way

Getting the length right

Sume does not accept a length field for music, so the 90 seconds are in the prompt and in your render. In Timeline, audio.duration_seconds is the narration length, and the voiceover is the spine. Pass the music as soundtrack.url with duck_db (a value from 0 to 20) so it dips while the voice speaks, and set fade_out_seconds for a soft ending.

A short track can be looped with soundtrack.loop. A long track is simply cut at the spine length. Because the spine is real audio here, ducking is allowed; it would be refused with a silent spine.

What it costs

Billing follows the output length, rounded up to started minutes. A 90-second render is two minutes, so trimming the script to 60 seconds halves the render line to $0.10. The music line does not change with length, since the price per accepted generation is fixed.

Check before you publish

Play the explainer to someone outside the clinic. If they remember the music, the bed is too present; if they remember the explanation, it is right.

Keep the job id and routed engine with the final file for your records.

Clinics often publish the same explainer in several lengths. For a 30-second version, change the prompt to a 30-second track and drop the midpoint brightening to 0:15; the render becomes one started minute at $0.10. For a three-minute patient guide, ask for a 3-minute track with a longer middle section. Each variant is its own generation at $0.125.

Staff and patients are a different audience from advertisers. A waiting-room screen can use a quieter bed with a higher duck_db, while a social cut can carry a little more energy. Write both briefs down so each is repeatable, and generate them separately instead of trying to make one track serve both.

• Do not use music to imply medical results; keep the claims in the script, not the bed.
• Check every spoken claim against what your clinic can support; Sume's tools do not do that review.
• Use POST /v1/timeline-1.0/plan first; it is unbilled and returns billable minutes.
• Confirm the usage terms for generated music with Sume before publishing a paid ad.