The AI Music Prompt Cheat Sheet: Every Word That Matters

After thousands of generations across every genre, a pattern emerges: the prompts that produce great songs share a vocabulary. The ones that produce flat songs are missing it.

This is the cheat sheet. Use it like a word bank.

Vocal Style

Be specific. "Male vocals" gets you something generic. "A weathered male baritone with gravelly texture and emotional weight" gets you a character.

Voice descriptors that work:

• Tenor, baritone, alto, soprano (range)
• Smooth, gravelly, breathy, raspy, smoky, crystalline, weathered, polished (texture)
• Powerful, restrained, intimate, conversational, soulful, controlled (delivery)
• With vibrato, with vocal fry, with a slight rasp (technique)

Phrases that lock in feel:

• "Whispered intimate vocals over deep sub-bass"
• "Gritty male baritone, almost spoken at times"
• "Bright female vocals with storytelling delivery"
• "Soulful tenor with controlled power"

Instrumentation

Don't just list instruments. Describe how they're played.

Weak: "guitar, drums, bass."

Strong: "Fingerpicked acoustic guitar, brushed snare, walking upright bass."

The difference: the second tells the model what kind of song this is. Picked vs. strummed acoustic = entirely different mood. Brushed vs. hit drums = jazz vs. rock.

Common pairings that work:

• Acoustic folk: fingerpicked acoustic, brushed snare, upright bass, harmonica, pedal steel
• Indie pop: clean electric guitar, programmed drums, synth pads, layered harmonies
• Country ballad: strummed acoustic, fiddle, dobro, soft pedal steel, brushed kit
• Lo-fi hip-hop: mellow drum machine, vinyl crackle, jazzy piano samples, warm bass
• Cinematic/orchestral: sweeping strings, brass swells, timpani, choral pads
• Electronic/EDM: sidechained synth bass, layered pads, four-on-the-floor kick, vocal chops

Mood Words That Hit

Stack two or three. One mood word is too thin.

Warm: nostalgic, tender, hopeful, bittersweet, comforting Energetic: anthemic, defiant, triumphant, celebratory, urgent Dark: moody, brooding, haunting, ominous, restless Soft: intimate, dreamy, contemplative, fragile, peaceful

Stack examples:

• "Warm, nostalgic, hopeful"
• "Defiant, triumphant, slightly bittersweet"
• "Moody, brooding, intimate"

Tempo Language

The model reads "BPM" but it also reads feel.

• Slow ballad: 60-80 BPM
• Mid-tempo groove: 90-110 BPM
• Upbeat pop/rock: 120-140 BPM
• Driving energy: 140+ BPM

Or skip the number and describe it: "unhurried", "loping", "steady mid-tempo", "driving", "frantic".

Production Language

This is the secret sauce most people miss. Production words shape the sound of the recording, not the song.

• Analog warmth, tape saturation — vintage feel
• Sparse arrangement, stripped-down — fewer instruments, more space
• Lush production, layered — many instruments, dense
• Front-porch recording, intimate mic — close, raw, no polish
• Wall-of-sound production — big, loud, full
• Spacious reverb, atmospheric — open, ambient
• Tight, dry mix — close, focused, no room sound

What to Avoid

Artist names. The model rejects prompts with copyrighted references. Translate the sound instead.

• ❌ "Sounds like Adele"
• ✅ "Powerful female vocals with piano-driven soul-pop, dramatic dynamic swells, intimate verses exploding into anthemic choruses"

Vague mood. "Happy" is too thin. Pick three specific feel words.

Lyric instructions. Don't write the lyrics. Describe the theme and let the model write them. (It rewrites supplied lyrics anyway.)

Contradictions. "Aggressive thrash metal acoustic ballad" confuses the model. Pick a lane.

The Formula

Combine the categories above into one dense paragraph:

[Genre/style descriptor]. [Vocal style]. [Instrumentation with playing details]. [Mood — stack two or three]. [Tempo feel]. [Production note]. [Theme — what the song is about].

That's it. Three to five sentences. Specific beats long.

Build your prompt with the assistant

Or paste your own. Three songs free, no credit card, results in under two minutes.