Instrumental AI music by API: how to ask for no vocals

For instrumental AI music, say so in the prompt: end it with "Instrumental, no vocals." Sume's Music 1.0 docs use exactly that closing clause, and Google's Lyria page says you can prompt the model for instrumental-only tracks by asking for no vocals. Sume's request has no vocals on/off field.

Sources: Sume's Music 1.0 and Music Router docs and Google's Lyria page, read 2026-09-29.

What should the prompt say for a background bed?

Name the mood, tempo, key and two to four instruments, then close with the exclusion. If the bed goes under narration, Sume's docs add "no spoken word".

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: bed-001" \
  -d '{
    "prompt": "Calm corporate ambient, 90 BPM, G major. Soft Rhodes, light pad, brushed shaker. A 1-minute track. Instrumental, no vocals, no spoken word."
  }'

Can I use negative_prompt instead?

No. Sume's docs say Lyria does not support negative prompting: a non-empty negative_prompt returns HTTP 400 with public_reason=negative_prompt_unsupported. Omit the field or send an empty string, and put every exclusion into the positive prompt.

How do I check that it is instrumental?

Listen. The docs treat prompt exclusions as directions, not guaranteed settings. The result.lyrics field, when present, is model-reported metadata, so an empty or missing lyrics is a hint but not proof. If a take has vocals, regenerate and state the exclusion again in the prompt.