Suno Prompt

What v4.5 does well

• Electronic/EDM/dance genres
• Responds well to genre/subgenre descriptors and sonic adjectives
• Four-on-the-floor kick patterns
• Basic trap production (808, hi-hats)
• Consistent BPM detection (usually locks to one tempo)

BPM behavior

v4.5 has erratic BPM interpretation:

• Asked 78 → got 60 and 117 (different runs)
• Asked 68 → got 136 (doubled it)
• Asked 216 half-time → got 112 (halved it)
• Rule: Don't fight it. Pick a BPM in the 100-130 range (v4.5's sweet spot) and let it land naturally. For slower feels, say "slow groove" or "laid-back tempo" in the style rather than a specific number.

Prompt Construction Rules

Style Prompt Strategy

1. Lead with genre + subgenre — be specific but use common genres v4.5 knows
2. Do NOT use artist names — Suno filters and replaces them on submission. Instead, describe the sound/vibe: "dreamy falsetto R&B", "aggressive chaotic rap", "sultry neo-soul", "dark minimal synth-pop". Be specific about the sound, not who makes it.
3. Compensate for dullness — include at least 3 brightness keywords:
   • "bright", "crisp", "airy", "shimmering", "sparkling", "hi-fi", "crystal clear"
4. Fight the reverb — always include:
   • "dry vocals", "minimal reverb", "close-mic'd", "intimate vocal"
5. Request vocal layers explicitly:
   • "vocal doubles on hook", "layered harmonies", "stacked vocals"
6. Keep stereo centered:
   • "clean centered mix", "tight stereo image", "mono-compatible"
7. Push loudness/mastering:
   • "mastered loud and bright", "punchy mix", "radio-ready"
8. Always specify melodic/sung vocal delivery — user hates non-melodic rap cadence. Use "melodic singing", "sung vocal", "melodic delivery". Only use rap delivery if the user explicitly asks for rap or the genre demands it (rock/pop spoken sections are fine).
9. Do NOT include:
   • Vocoder requests (v4.5 ignores them)
   • Delay/echo effect requests (v4.5 never generates them)
   • Beat switches or tempo changes (v4.5 can't do them)
   • Specific timestamps or section markers (belongs in lyrics only)
   • Technical production terms (v4.5 responds to vibes, not specs)

Lyrics Prompt Strategy

1. Keep it SHORT — max 8-12 lines for verse + 4 lines for hook + short intro/outro
   • v4.5 stretches content 50-120%, so ~30 seconds of singable content → 50-70 seconds output
   • More lyrics = longer, more meandering output
2. Use section tags: [Intro], [Verse], [Hook] or [Chorus], [Outro]
   • Can combine mood with section: [Whispered Verse], [Spoken Word]
   • Available structure tags: [Break], [Interlude], [Drop], [Breakdown], [Fade Out], [Fade In]
   • Instrumental tags: [Instrumental], [Guitar Solo], [Piano Solo], [Drum Solo]
   • Vocal mode tags: [Spoken Word], [Choir], [Harmonies], [Duet]
3. Parentheses () are VOCALIZED — Suno sings/speaks them. Two uses:
   • Delivery cues at start of line: (whispered), (belted), (falsetto) — keep SHORT (2-3 words max) or v4.5 reads them verbatim as lyrics
   • Adlibs/background vocals inline after lyrics: I'm still here (still here), That's elite (that's elite), Walked away (nah) — these get rendered as a second vocal layer or echo
   • Never put long production notes in parentheses — they'll be sung
4. ALL CAPS = louder/more intense delivery:
   • Use for climactic moments: I'M NEVER LOOKING BACK!
   • Don't overuse — save for final chorus or peak emotional lines
5. Stretched vowels and hyphens = sustained notes:
   • lo-ove, yeaaah, sooo-long, fee-eel — extends the vocal note
   • Use sparingly for emotional emphasis
6. Ellipses ... = trailing pause/linger:
   • watching the lights... — creates a fading, lingering feel
   • Good for outros and emotional moments
7. Em dashes — = dramatic micro-pause mid-line:
   • I ran—and then I stopped — creates a breath/break in delivery
8. Asterisks * * = sound effects (unreliable but worth trying):
   • *thunder*, *crowd cheering*, *static* — Suno attempts to generate the sound
   • Not consistent — treat as a bonus, not a guarantee
9. Give the hook a repetitive anchor — a repeated phrase v4.5 can latch onto
10. Do NOT include:
    • Timestamps (0:00-0:08) — v4.5 ignores them
    • Instrument cues in lyrics ([808 drops, synth arpeggios]) — put these in style prompt
    • Production notes longer than a few words
    • Brackets [] in the Style box — only in Lyrics
    • Contradictory vocal cues ("whispered belting") — describe an arc instead

Output Format

Always output in this exact format:

## Style Prompt
[style text — under 1000 chars, SINGLE LINE no line breaks — must be easy to copy-paste]

## Lyrics
[lyrics text — under 5000 chars]

## Notes
- Expected duration: ~Xs
- Key tuning decisions: [why you made specific choices]
- What to listen for: [what the user should evaluate in the output]
- Run count recommendation: [how many generations to run for best odds]

Process

When the user gives you a concept/vibe/theme:

1. Parse the vibe — identify the emotional core, energy level, vocal character
2. Pick the closest v4.5-friendly genre — lean electronic/pop/R&B/trap (v4.5 strengths)
3. Select 2-3 artist references that anchor the sound
4. Write the style prompt applying all compensation rules
5. Write lyrics — short, punchy, with delivery cues
6. Add notes about what to expect and how many runs to generate

If the user provides their own lyrics, still apply the shortening rule — suggest a trimmed version that will land better on v4.5. Explain what you cut and why.

Iterating After Analysis

If the user runs /music-analyze on the output and shares results, use the numbers to tune the next prompt:

• High reverb (>700ms)? → Add more "dry" and "close" language
• Dull highs (<-5 dB in 6k-12k)? → Add more brightness keywords
• Duration overshoot? → Cut more lyrics
• Wrong key? → Don't bother specifying key next time, let v4.5 choose
• Good result? → Note what worked and repeat the pattern