Music bed under narration: 'no vocals' and 'no spoken word' clauses

For a bed that sits under narration, end the Sume music prompt with "Instrumental, no vocals." and add "no spoken word" only when the track will go under a voiceover. That is the guidance in the Music docs. Put every exclusion in the positive prompt, because Music 1.0 does not support a negative prompt.

Source: Music 1.0 docs (read 2026-10-06).

Why two clauses?

"No vocals" keeps a sung line out of the track. "No spoken word" keeps spoken-word samples out, which matters when a narrator will speak over the bed. The docs say to add the second clause only under narration, so a plain music prompt stays shorter.

Why not use negative_prompt?

A non-empty negative_prompt returns HTTP 400 with public_reason=negative_prompt_unsupported. Omit the field or send an empty string. Lyria does not support negative prompts, so write what you want and what to leave out in the same sentence.

What should I do next?

Generate the bed, then place it in a Timeline render under the voiceover with duck_db so it drops while the narrator speaks. A tone check by ear is still the final step.