Why every AI background track sounds the same: a seven-axis brief

AI background tracks sound alike when the prompts are alike. Sume's Music docs ask for a scene-specific brief with seven axes: emotion, genre, tempo as a number, key and mode, two to four instruments with texture, an arc with one named moment, and an era or production note. A prompt that names only a mood leaves every other choice to the model's defaults.

The axes and the example come from the Music 1.0 docs (read 2026-10-06).

What are the seven axes?

The axes are creative directions, not guaranteed output values, so check the audio you get.

How do I make a batch contrast?

For scenes in one project that should differ, change the broad genre family, the tempo by at least 12 BPM, and the lead instrument. If you want one consistent score, keep those fixed. You can also pass a scene still as image_url.

What else do I add?

End with one clause such as "Instrumental, no vocals." Put exclusions in the positive prompt, because a non-empty negative_prompt returns HTTP 400. The prompt can run up to 5,000 characters.