Why instrumental only
Say "instrumental" directly in the prompt, since square-bracket tags such as [Instrumental] are a common convention that models follow often but not always. Vocals compete with speech for attention, so even a wordless "la la" hook can distract from narration. If the model adds a voice anyway, add "no vocals, no vocal melody" to the descriptors and try again.
Why lo-fi works for most video backgrounds
Lo-fi (relaxed, slightly worn-sounding hip-hop-influenced instrumental music) sits well under speech because its drums are soft, its bass is rounded rather than sharp, and its melodies are simple chord loops rather than a strong hook competing for the ear. Typical tempo is 60 to 90 BPM, most often 70 to 85 BPM, in keys like D minor or C major with seventh and ninth chords (added notes that give a mellow, jazzy flavour without extra loudness). See the lo-fi prompt page for more on this style.
What to avoid describing
- Skip words like "epic", "cinematic build" or "drop" — these invite loud sections that clash with talking.
- Skip strong lead instruments (saxophone solo, lead guitar riff) since they pull focus from the voice.
- Skip fast tempo changes; a background track should hold one steady pace.
- Skip quality words such as "masterpiece" or "best" — in our measurements on ACE-Step 1.5 these did not change the sound.
How Lyro Music handles this
Lyro Music turns a description into a mixed and mastered instrumental track. Since these prompts have no lyrics, the render routes through ACE-Step 1.5, which also handles tempo-locked beats. You can start from the free prompt generator at /ai-music-prompt-generator, which builds a prompt like the ones above from a plain description, and tracks run 30 seconds to 3 minutes, useful for looping under a short clip or a longer video.
Common mistakes
- Adding a strong vocal or hook melody: fix by naming "instrumental" and "no lead melody" so nothing competes with speech.
- Asking for cinematic builds or drops: fix by keeping mood words steady, like "calm" or "neutral", not "epic" or "intense".
- Piling on too many moods at once (calm, upbeat, dramatic): fix by picking one mood and one tempo range that matches the footage.
- Using dB or percentage numbers to control the mix: fix by describing texture instead, like "soft" or "warm", since numbers read as placebo in our measurements.
- Forgetting tempo: fix by naming a BPM range (70 to 90 for most background use) so the pace matches editing cuts.