How to Write AI Song Prompts

Prompt formula

Start with result, sound, emotion, and limits

The most reliable prompt starts by naming the output, then the sound, then the feeling, then a few constraints. This gives MemoTune a target instead of a loose mood board. A practical formula is: song type + genre + mood + instruments + rhythm + voice + story + production limit. You do not need every field for every song, but you should know which field you are leaving out and why.

• Output: full song, intro, jingle, instrumental, verse
• Sound: genre, instruments, era, vocal type
• Emotion: hopeful, cinematic, playful, intimate
• Limits: clean vocals, no spoken intro, avoid harsh drums

Specificity

Replace vague adjectives with musical clues

Words like cool, catchy, or professional are too broad on their own. Pair them with concrete clues such as 90 BPM, acoustic guitar, female lead vocal, minor key, or radio-ready chorus. If you want a cheerful track, say whether the cheer comes from claps, bright synths, major-key piano, dance drums, or a smiling vocal delivery. If you want a dark track, say whether the darkness is cinematic, industrial, minimal, gothic, or trap-influenced.

Rewrite pattern

Improve weak prompts by adding missing dimensions

If a result feels generic, revise the prompt by adding one missing dimension at a time: style, vocal, story, arrangement, tempo, or negative instruction. A weak prompt says “make a sad love song.” A stronger version says “slow indie pop love song with soft male vocals, fingerpicked acoustic guitar, rainy-night mood, lyrics about apologizing after a breakup, gentle chorus, no spoken intro.” The improvement comes from concrete musical decisions, not simply from adding length.

Tempo and key

Use tempo, key, and energy only when they help

Tempo and key are useful when you want predictable motion. A workout song may need 128 BPM, a lullaby may need a slow pulse, and a cinematic suspense cue may need a gradual build instead of a fixed beat. Key language can also help mood: minor keys often feel darker, while major keys often feel brighter. If you are not sure, describe energy in plain language: slow, mid-tempo, fast, half-time, driving, floating, or steadily building.

• Fast: club, workout, chase scene, celebration
• Mid-tempo: pop storytelling, podcast intro, casual groove
• Slow: ballad, ambient, lullaby, reflective lyric
• Build: cinematic reveal, chorus lift, trailer cue

Voice direction

Describe the singer as a performance, not just a gender

Vocal direction should explain the performance. “Female vocal” or “male vocal” is a start, but “warm female vocal with intimate delivery” or “raspy male vocal with restrained emotion” gives more useful guidance. You can also describe choir, duet, spoken narration, soft falsetto, confident rap flow, layered harmonies, children’s choir, or call-and-response. If the song is instrumental, say “no vocals” and spend the saved space on instruments, groove, and use case.

Avoid tags

Add one or two avoid instructions

Avoid instructions keep the output from drifting into unwanted territory. They work best when they are short and specific: no spoken intro, no harsh drums, no long fade, no choir, no rap verse, no heavy distortion, no comic sound effects, or no vocals. Too many negatives can make a prompt confusing, so choose the two boundaries that matter most for the song’s purpose.

Structure cues

Use bracketed cues when lyrics need structure

When writing in lyrics mode, bracketed cues can separate sections and performance notes from sung words. Use labels such as [Verse 1], [Chorus], [Bridge], [Outro], [Spoken Intro], or [Fade Out] when the structure matters. Keep those cues concise. The lyric lines should still carry the story, while the tags tell MemoTune how to organize the performance.