By name: suno-lyric-writer description: >- Use when writing, reviewing, or revising song lyrics for Suno AI generation. Covers the full pipeline: lyric drafting with professional prosody and rhyme craft, Suno pronunciation scanning (homographs, tech terms, proper nouns, acronyms), 14-point quality review, and complete Suno v5/v5.5 prompt engineering including Section Tags, Metatags, Vocal-Delivery Tags, Persona/Voice/Custom-Model workflows, Creative Sliders, Extend/Cover/Remaster strategies and Negative Prompting. Triggers on: write lyrics, song text, Suno track, lyric review, let's work on a track, new song, revise lyrics, lyric QC, prosody check, pronunciation scan, review my lyrics, check lyrics for Suno, songwriting, write a song, Suno prompt, style prompt, suno tags, suno voice, persona, suno extend, suno cover, suno remaster, suno sliders. metadata: category: creative source: adapted from bitwize-music-studio/claude-ai-music-skills (Apache 2.0) date_added: "2026-04-08" version: "2.0.0" triggers: >- lyrics, songwriting, Suno, track, prosody, rhyme, write a song, lyric review, pronunciation, homograph, review lyrics, QC, check lyrics, suno prompt, style tags, vocal tags, persona, voice, extend, cover, remaster, sliders
Suno Lyric Writer (v5 / v5.5)
You are a professional lyric writer, pronunciation specialist, and QC reviewer — three roles unified into one pipeline optimized for Suno AI v5 / v5.5.
Supporting Files (read when referenced)
| File | Contents | When to Read |
|---|---|---|
| craft-reference.md | Rhyme types, section length tables, density rules, refinement pass patterns | Before drafting; during QC checks 8–12 |
| pronunciation-guide.md | Homograph table, tech terms, names, acronyms, numbers, scanning rules | During pronunciation scan (Phase 2) |
| documentary-standards.md | Legal standards for lyrics about real people/events | Only if lyrics reference real people/events |
| examples.md | Before/after transformations for all key principles | When explaining fixes or demonstrating craft |
| genre-practices.md | Genre-specific Suno prompt strategies, voice types, production direction | During Phase 4 (Suno prompt engineering) |
Unified Pipeline
Five phases, executed in sequence for every lyric task:
Phase 1: WRITE Phase 2: PRONUNCIATION Phase 3: QC REVIEW Phase 4: SUNO PROMPT Phase 5: WORKFLOW
──────────── ────────────────────── ────────────────── ────────────────── ──────────────────
Draft lyrics Scan for risky words 14-point checklist Build style prompt Persona / Voice
Core principles Flag homographs (ASK) Severity report Tags + Metatags Sliders
Refinement passes Auto-fix tech/acronyms Ready-for-Suno gate Negative prompting Extend / Remaster
Phase 1: Write Lyrics
Input Gathering
If any of these are missing, ask before drafting:
- Genre (determines rhyme scheme, section limits, density)
- Mood / Theme (drives imagery, vocabulary, energy)
- Target duration (default: 3:30–5:00 → see craft-reference.md for word count mapping)
- BPM (determines max verse density — see craft-reference.md BPM table)
- POV (first/second/third person)
- Concept album context? (if yes: which track number, previous track themes for cross-referencing)
Core Principles
Rhyme Craft
- Never rhyme a word with itself; no near-repeats (mind/mind, time/time)
- No lazy predictable pairs (fire/desire, moon/June, night/light)
- Use variety: perfect, slant, consonance, assonance, internal (see craft-reference.md)
- Meaning over rhyme — if a perfect rhyme sounds unnatural, use a near rhyme
Prosody (Syllable Stress)
- Stressed syllables land on downbeats (beats 1 and 3 in 4/4)
- Multi-syllable words need natural emphasis: DES-ti-ny, not des-TIN-y
- Content words (nouns, verbs) take emphasis, not function words (the, into, a)
- Syllable balance: V1 and V2 must match within ±2 syllables per line — drift causes timing collapse
- Test: Speak the lyric aloud. If emphasis feels wrong, rewrite it.
Show Don't Tell
- ACTION:
❌ "My heart is breaking"→✅ "She fell to her knees as he packed his bag" - IMAGERY:
❌ "I felt so sad"→✅ "Coffee gone cold on the counter" - SENSORY DETAIL: Engage multiple senses (sight, sound, smell, touch, taste, kinesthetic)
- Section balance: Verses = sensory details. Choruses = emotional statements.
Verse/Chorus Contrast
| Element | Verse | Chorus |
|---|---|---|
| Lyrics | Observational, narrative | Emotional, universal |
| Energy | Building | Peak |
| Detail | Specific sensory | Abstract emotional |
No Verse-Chorus Echo
A verse must never repeat a key phrase, image, or rhyme word from the chorus it leads into.
Hook & Title
- Title in first or last line of chorus; give it rhythmic accent and melodic peak priority
V2 Must Develop (No Twin Verses)
V2 must advance the story, deepen emotion, or shift perspective — never rephrase V1.
Section & Length Limits
Refer to craft-reference.md for genre-specific tables. Universal rules:
- Max words: 400 (non-hip-hop), 600 (hip-hop). Hard fail above.
- Min words: 200 for 3:30+ tracks.
- Add more sections, not longer sections to hit duration targets.
- Instrumental tags add ~20–40 seconds each.
- Max lines per section: 6–8 (v5 processes up to ~12 syllables/line cleanly)
No Invented Contractions
Suno only handles standard contractions (they'd, he'd, wouldn't).
❌ signal'd, TV'd → ✅ "signal would", "TV could"
Refinement Passes
After drafting, run 1 pass (configurable 0–3):
| Pass | Focus | Goal |
|---|---|---|
| 1 — Tighten | Cut filler, compress, remove redundancy | Every word earns its place |
| 2 — Strengthen | Upgrade weak imagery, sharpen sensory detail | Lines that stick |
| 3 — Flow & Ear | Read-aloud test, singability at BPM | Sounds right when sung |
Phase 2: Pronunciation Scan
One wrong word ruins the take. Suno guesses pronunciation — wrong guess = wasted generation.
| Category | Risk | Action |
|---|---|---|
| Homographs (live, read, lead, wind, tear, bass) | CRITICAL | Always ASK user — never guess |
| Tech terms (Linux, SQL, API) | High | Auto-fix with phonetic |
| Names / proper nouns | High | Phonetic spelling mandatory |
| Acronyms (FBI, GPS) | Medium | Spell out with hyphens |
| Numbers (1993, 404) | Medium | Write out or use apostrophes |
Read pronunciation-guide.md for complete tables. Generate a Pronunciation Notes Table for any track with risky words:
| Word/Phrase | Phonetic | Notes |
|-------------|----------|-------|
| live | lyve | Verb — user confirmed |
| SQL | S-Q-L | Auto-fix |
Every entry MUST be applied as phonetic spelling in the final Suno lyrics.
Phase 3: QC Review (14-Point Checklist)
| # | Check | Severity | What to Scan |
|---|---|---|---|
| 1 | Rhyme | ⚠ | Self-rhymes, repeated end words, lazy patterns |
| 2 | Prosody | ⚠ | Stress misalignment, inverted word order |
| 3 | Pronunciation | 🔴 | Homographs unresolved, proper nouns unphonetic |
| 4 | POV/Tense | ⚠ | Inconsistent pronouns or tense within section |
| 5 | Structure | ⚠ | Missing tags, twin verses, buried hook |
| 6 | Flow | ⚠ | Forced rhymes, inverted word order, filler |
| 7 | Documentary | 🔴 | Internal state claims, fabricated quotes (conditional) |
| 8 | Factual | 🔴 | Wrong dates/names/facts (conditional) |
| 9 | Length | 🔴 | Word count vs genre target |
| 10 | Section length | 🔴 | Lines per section vs genre max |
| 11 | Rhyme scheme | ⚠ | Scheme matches genre, no orphan lines |
| 12 | Density/pacing | 🔴 | Verse lines vs BPM-aware limits |
| 13 | Verse-chorus echo | ⚠ | Shared phrases/images across boundaries |
| 14 | Artist names | 🔴 | Real artist names in lyrics or style prompt |
Severity & Report Format
| Level | Meaning | Action |
|---|---|---|
| 🔴 Critical | Suno will fail or mispronounce | Must fix before generation |
| ⚠ Warning | Quality issue | Should fix, can proceed |
## Lyric Review — [Track Title]
**Genre**: [genre] | **BPM**: [bpm] | **Words**: [count]
**Status**: ✅ Ready / ❌ Needs Fixes
### 🔴 Critical Issues
- [#3] V1:L2 — "lead" homograph, unresolved
### ⚠ Warnings
- [#1] V2:L2-L4 — self-rhyme "night/night"
### ✅ Auto-Fixed
- "FBI" → "F-B-I" (V2:L3)
### Ready-for-Suno Gate
- [ ] Zero critical issues - [ ] All pronunciation notes applied
- [ ] No unresolved homographs - [ ] Word count within genre target
Phase 4: Suno Prompt Engineering (v5 / v5.5)
V5 Core Rules
- V5 is literal — simple, direct prompts > verbose descriptions
- Vocals FIRST in style prompt — always lead with vocal description
- Section tags are critical — v5 uses them to shape arrangement
- Suno sings EVERYTHING in the lyrics box — no production notes inline
- Max 2 genres — 3+ produces inconsistent results
- Style Prompt limit: 1.000 chars | Lyrics limit: 5.000 chars
- BPM and key are reliable in v5 — include both when precision matters
Lyrics Box Format
[Intro]
[Verse 1]
First line of lyrics
(backing ad-lib in parentheses)
[Pre-Chorus] / [Chorus] / [Post-Chorus]
[Instrumental Break]
[Verse 2] / [Bridge]
[Final Chorus]
[Outro]
[End]
Complete Section Tag Reference
Core: [Intro] [Verse 1] [Pre-Chorus] [Chorus] [Post-Chorus] [Bridge] [Interlude] [Break] [Hook] [Refrain] [Outro] [End]
Instrumental: [Instrumental] [Solo] [Guitar Solo] [Breakdown] [Drop]
V5 Dynamics (new): [Build] [Build-Up] [Final Chorus] [Fade In] [Fade Out] [Swell] [Crescendo] [Decrescendo]
Formale Kategorie-Metatags (granulare Kontrolle, v5-spezifisch):
[Mood: Uplifting] [Mood: Introspective] [Energy: High] [Energy: Medium→High]
[Instrument: Warm Rhodes, Soft Drums] [Texture: Gritty]
[Vocal Style: Whisper] [Vocal Style: Raspy] [Structure: seamless loop]
Vocal-Delivery-Tags (inline, vor oder innerhalb Sektionen)
| Lautstärke | Stil | Techniken | Emotion |
|---|---|---|---|
[Whispered] [Soft] [Spoken] |
[Falsetto] [Breathy] [Raspy] |
[Harmonies] [Ad-libs] [Melisma] |
[Vulnerable] [Defiant] [Sultry] |
[Powerful] [Belted] [Screamed] |
[Smooth] [Soulful] [Operatic] |
[Vibrato] [Choir] [Call and Response] |
[Melancholic] [Joyful] |
Rap: [Rapped] [Fast Rap] [Double Time] [Trap Flow] [Boom Bap Flow]
Effects: [No AutoTune] [Vocoder] [Telephone Effect] [Distorted Vocals]
Inline-Beispiel:
[Verse 1]
[Whispered] In the silence of the night
[Building] I feel you close to me
[Belted] AND I CAN'T LET GO!
Lyrics-Box-Tricks:
(oh yeah)Runde Klammern → Ad-Lib / Backing-Vocal Layer(*synth swirl*)Asterisken in Parenthesen → diskrete Produktions-Cues- Emphasis:
loooove,feeeel; Silbentrennung:lo-ove - Duett:
[Male Vocal] line/[Female Vocal] line/[Duet] line
Style Prompt Construction
Formel: [VOCAL] [GENRE(S) + ÄRA] [2–3 INSTRUMENTE] [PRODUKTION + BPM + KEY]
- Top-Loading: Genre und Stimmung zuerst — v5 gewichtet erste Wörter am stärksten
- 4–8 Tags Sweet Spot; Ankertechnik: Schlüsseldeskriptoren Anfang UND Ende
- Ärabeschreibungen statt Artist-Namen: „late 70s disco", „80s goth"
Female alto, haunting breathy vocals. Darkwave, synth goth, atmospheric analog synths,
driving bass, programmed drums. Dark, spacious reverb. 140 BPM, D minor.
no reverb lead, no electric guitar.
Artist-Name-Ersatz-Tabelle:
| Don't Write | Write Instead |
|---|---|
| "Depeche Mode" | "dark synth-pop, brooding male vocals, analog synths" |
| "NIN" | "dark industrial, grinding synths, distorted vocals" |
| "Siouxsie" | "post-punk goth, commanding female vocals, jangly guitar" |
| "Sisters of Mercy" | "goth rock, deep baritone, drum machine, atmospheric" |
| "Massive Attack" | "trip-hop, dark atmospheric, sparse beats, cinematic" |
Exclude Styles (Negative Prompting)
no [Element] am Ende des Style Prompts — max 4–5 Items. Syntax-Varianten:
no drums no autotune no electric guitar no choir no sidechain pump
no heavy compression no over-mastered sound raw recording feel
Ghost-Vocal-Dreifach-Sicherung für Instrumentals: UI-Toggle + no vocals, no singing, no humming, no choir + [Instrumental] in jeder Lyrics-Sektion.
Duration Awareness
| Target | Structure Guidance |
|---|---|
| < 2:00 | 1–2 sections + [End]. Add "short" in style prompt |
| 2:00–3:00 | 2 verses max, short bridge |
| 3:00–5:00 | Standard (v5 generiert bis zu 4 Min. in einem Pass) |
| 5:00+ | 3+ verses, pre-chorus, bridge, 1–2 instrumental breaks. Add "extended" |
Voice & Delivery Quick Reference
| Type | Range | Best For |
|---|---|---|
| Soprano | High female | Pop, theatrical |
| Alto | Low female | Jazz, darkwave, folk |
| Tenor | High male | Pop, rock, R&B |
| Baritone | Mid male | Rock, goth, country |
| Bass | Low male | Blues, doom |
Phase 5: V5/V5.5 Workflow-Features
Creative Sliders
| Slider | Funktion | Empfohlener Bereich |
|---|---|---|
| Weirdness | Unvorhersagbarkeit/Experimentierfreude | 35–55 standard; Chorus: tief, Bridge: höher |
| Style Influence | Strenge der Prompt-Befolgung | 55–80 für Genre-Treue |
| Audio Influence | Treue zum Upload-Material | 25–40 Textur; 60–75 enge Stimm-Matches |
Immer nur einen Slider gleichzeitig ändern, 2–4 Varianten pro Änderung generieren.
Personas, Voices (v5.5) & Custom Models
Personas — Stimme + Style eines Songs als Template speichern:
- ⋮ → Create → Make Persona. In Custom Mode auswählen.
- Aus Tracks mit spärlicher Instrumentierung und trockenen Vocals ableiten.
- Bei Persona: Style Prompt vereinfachen (Persona trägt die Identität).
Voices (v5.5, Pro/Premier) — Voice Cloning:
- Upload/Aufnahme mit Anti-Deepfake-Verifizierung; max 3 pro Account.
- Voice Influence bei 50% starten (höhere Werte = Shimmer-Artefakte).
- Bei Voices: Gender-Deskriptoren aus Prompt entfernen.
Custom Models (v5.5) — eigener Produktionsstil via eigene Tracks (max 3).
- Konsistenter Stil im Katalog → bessere Ergebnisse. Gemischte Genres degradieren das Model.
Extend, Cover, Remaster, Stems
Extend: Von Momentum-Punkten aus (Mitte Strophe, nicht nach finalem Chorus).
[Callback: continue with same vibe as chorus] verhindert Style-Drift.
Fertige Sektionen aus Lyrics löschen — sonst Wiederholung statt Fortsetzung.
Cover: Weirdness 0–30% = nah am Original; 70–100% = radikale Neuinterpretation.
Remaster: Subtle / Normal / High. Alte v3.5/v4-Tracks sofort auf v5-Qualität.
Stems & DAW: Bis zu 12 Stems-Export. Tempo-Drift-Fix: Studio Transport-Bar → Manual BPM setzen vor Export.
Anti-AI-Sound: organic feel, human performance, no quantization, subtle imperfections, live recording vibe
Common Issues & Fixes (v5)
| Problem | Fix |
|---|---|
| Vocals buried | Vocal-Beschreibung ZUERST im Style Prompt |
| Wrong genre | Spezifischere Deskriptoren, max 2 Genres |
| Song cuts off | [Outro] + [End] Tags |
| Repeating sections | Klare Section-Tags, V2 variiert |
| Mispronunciation | Phonetic spelling in Lyrics-Box |
| Unwanted elements | Exclude Styles (max 4–5 Items) |
| Ghost-vocals (instrumental) | Dreifach-Sicherung (s. oben) |
| Style drift in extensions | [Callback] Tag + fertige Sektionen löschen |
| AI-Sound zu glatt | raw recording feel, no heavy compression, subtle imperfections |
| Tempo-Drift in DAW | Manual BPM im Studio setzen |
| Wrong BPM | Explizit in Style Prompt: 140 BPM |
Deliverables
- Final lyrics with Suno section tags + phonetic fixes applied
- Pronunciation Notes table (if risky words exist)
- 14-point QC Report with severity grades
- Refinement Log (before→after changes)
- Suno Input Package:
- Style of Music (vocals first → genre → instruments → production → BPM/Key)
- Exclude Styles (optional, max 4–5 items)
- Lyrics Box (clean lyrics + tags + phonetics)
- Settings (v5/v5.5, Instrumental, Persona/Voice if applicable)
- Creative Sliders recommendation (Weirdness / Style Influence)
- Streaming lyrics (standard spelling, no phonetics) — if requested
Pitfalls Checklist
- Forced emphasis on wrong beats | Inverted word order for rhyme
- Predictable rhymes (moon/June, fire/desire) | POV or tense inconsistency
- Twin verses (V2 rephrases V1) | No hook / buried title
- Section exceeds genre max | Orphan lines / broken rhyme scheme
- Filler phrases | 8-line verse at BPM under 100 | 3+ proper nouns in a verse
- Verse-chorus echo | Invented contractions
- V5-specific: V1/V2 syllable count mismatch (>±2) | >2 genres in Style Prompt
- Artist names anywhere | Structure tags in Style Prompt (belong in Lyrics Box)
Cross-Track Referencing (Concept Albums)
For track N>1: review previous tracks → identify 1–3 callback opportunities → weave organically. Types: callback, motif, character thread, contrast/inversion, resolution. Bookend rule: Final track echoes track 1. Quality: Subtle > heavy. Track must stand alone first.