pronunciation-specialist
À propos
Cette compétence analyse les paroles de chansons pour détecter les risques de prononciation, tels que les noms propres, les termes techniques et les mots non-anglais, afin d'éviter les erreurs de prononciation dans la génération musicale Suno AI. Elle examine les paroles fournies ou lit les fichiers audio (en ignorant les instrumentaux) et signale les problèmes avec des corrections suggérées. Utilisez-la lors de l'écriture de paroles contenant des éléments que la synthèse vocale automatisée pourrait prononcer incorrectement.
Installation rapide
Claude Code
Recommandénpx skills add bitwize-music-studio/claude-ai-music-skills -a claude-code/plugin add https://github.com/bitwize-music-studio/claude-ai-music-skillsgit clone https://github.com/bitwize-music-studio/claude-ai-music-skills.git ~/.claude/skills/pronunciation-specialistCopiez et collez cette commande dans Claude Code pour installer cette compétence
Documentation
Your Task
Input: $ARGUMENTS
Instrumental Guard
When invoked with a track file path, first check the track's frontmatter for instrumental: true or the Track Details table for **Instrumental** | Yes. If the track is instrumental:
- STOP and report: "SKIP — Instrumental track (no lyrics to scan for pronunciation)"
- Do NOT scan instrumental tracks.
Vocal Track Workflow
Based on the argument provided:
- If given a track file path: Read it, scan lyrics for pronunciation risks, report issues with fixes
- If given lyrics directly: Scan and flag risky words
- Output: Clean lyrics with all phonetic fixes applied, ready for suno-engineer
Supporting Files
- word-lists.md - Complete tables of homographs, tech terms, names, acronyms, numbers
Pronunciation Specialist
Scan lyrics for pronunciation risks, suggest phonetic spellings, prevent Suno mispronunciations.
Why This Matters
The problem: Suno AI guesses pronunciation. Wrong guess = wrong song = wasted generation.
One wrong word ruins the take.
When to Invoke
Always invoke between lyric-writer and lyric-reviewer:
lyric-writer (WRITES + SUNO PROMPT) → pronunciation-specialist (RESOLVES) → lyric-reviewer (VERIFIES) → pre-generation-check
|
Scan, resolve, fix risky words
Your role — RESOLVE:
- The lyric-writer flags potential pronunciation risks and asks about homographs
- You do the deep scan, resolve ambiguities with the user, and apply all phonetic fixes
- The lyric-reviewer then verifies all resolutions were correctly applied
High-Risk Word Categories
See word-lists.md for complete tables. Summary:
1. Homographs (CRITICAL)
Same spelling, different pronunciation. ALWAYS require clarification.
(Canonical reference: ${CLAUDE_PLUGIN_ROOT}/reference/suno/pronunciation-guide.md. Keep this summary in sync.)
| Word | Options | Fix |
|---|---|---|
| live | LYVE (verb) / LIV (adjective) | "lyve" or "liv" |
| read | REED (present) / RED (past) | "reed" or "red" |
| lead | LEED (guide) / LED (metal) | "leed" or "led" |
| wind | WYND (air) / WINED (coil) | "wynd" or "wined" |
| tear | TEER (cry) / TARE (rip) | "teer" or "tare" |
| bass | BAYSS (music) / BASS (fish) | "bayss" or "bass" |
2. Tech Terms
Suno often mispronounces tech words:
- Linux → "Lin-ucks" (not "Line-ucks")
- SQL → "S-Q-L" or "sequel"
- API, CLI, SSH → spell out with hyphens
3. Names & Proper Nouns
Non-English names need phonetic spelling:
- Jose → "Ho-zay"
- Ramos → "Rah-mohs"
- Sinaloa → "Sin-ah-lo-ah"
4. Acronyms
3-letter acronyms → spell out with hyphens (FBI → F-B-I) Word-like acronyms → phonetic (RICO → Ree-koh, NASA → Nah-sah)
5. Numbers
- Years: Use apostrophes ('93) or words (nineteen ninety-three)
- Digits: Write out (four-oh-four, not 404)
Pronunciation Guides
You reference TWO pronunciation guides:
Base Guide (Plugin-Maintained)
- Location:
${CLAUDE_PLUGIN_ROOT}/reference/suno/pronunciation-guide.md - Contains: Universal pronunciation rules, common homographs, tech terms
- Updated: By plugin maintainers when new issues are discovered
Override Support
Check for custom pronunciation entries:
Loading Override
- Call
load_override("pronunciation-guide.md")— returns override content if found (auto-resolves path from config) - If found: load and merge with base guide (override entries take precedence)
- If not found: use base guide only (skip silently)
Override File Format
{overrides}/pronunciation-guide.md:
# Pronunciation Guide (Override)
## Artist Names
| Name | Pronunciation | Notes |
|------|---------------|-------|
| Ramos | Rah-mohs | Character name |
## Album-Specific Terms
| Term | Pronunciation | Notes |
|------|---------------|-------|
| Sinaloa | Sin-ah-lo-ah | Location |
How to Use Override
- Add artist names, album-specific terms, and genre-specific jargon
- Override entries take precedence over base guide entries for the same word
- Base guide updates via plugin updates without conflicts
- Override guide is version-controlled with your music content
Scanning Workflow
Step 1: Automated Scan via MCP
- Extract lyrics:
extract_section(album_slug, track_slug, "lyrics") - Homograph scan:
check_homographs(lyrics_text)— returns found homographs with line numbers, pronunciation options - Additional manual scan for tech terms, acronyms, numbers, and names (not covered by MCP homograph list) — cross-reference word-lists.md
- If style prompt exists:
scan_artist_names(style_text)— catch blocklisted names
After fixes are applied:
5. Verify: check_pronunciation_enforcement(album_slug, track_slug) — confirms all pronunciation table entries appear in lyrics
Step 2: Review Results
From MCP results and manual scan:
- Which words were flagged?
- What's the recommended fix for each?
Step 3: Generate Report
For each flagged word, provide:
- Line number and context
- Why it's risky (ambiguity type)
- Suggested phonetic spelling
- Alternative if multiple pronunciations exist
Example output:
PRONUNCIATION RISKS FOUND (3):
Line V1:3 -> "We live in darknet spaces"
Risk: "live" is homograph
Options: "lyve" (verb) or "liv" (adjective)
-> Needs clarification
Line C:1 -> "SQL injection in the code"
Risk: "SQL" is tech acronym
Fix: "S-Q-L" or "sequel"
-> Auto-fix: "S-Q-L injection in the code"
Line V2:5 -> "Reading Linux logs at 3AM"
Risk: "Linux" commonly mispronounced
Fix: "Lin-ucks"
-> Auto-fix: "Reading Lin-ucks logs at 3 A-M"
Step 4: User Confirmation
For ambiguous words (like "live"): Ask user which pronunciation For clear fixes (tech terms): Auto-fix
Auto-Fix Rules
Always Auto-Fix
- Tech terms (SQL → S-Q-L, Linux → Lin-ucks)
- Common acronyms (FBI → F-B-I, GPS → G-P-S)
- Numbers (1993 → '93 or nineteen ninety-three)
Ask User First
- Homographs (live, read, lead, wind, tear)
- Names (confirm pronunciation preference)
- Words with regional variants (data, either, route)
Output Format
Track File Updates
If given a track file, update these sections:
Pronunciation Notes (add table):
| Word/Phrase | Phonetic | Notes |
|-------------|----------|-------|
| Jose Diaz | Ho-say Dee-ahz | Spanish name |
| live | lyve | Verb form (to reside) |
| SQL | S-Q-L | Spell out |
Lyrics Box (apply fixes): Replace standard spelling with phonetic in the Suno lyrics section.
Standalone Report
PRONUNCIATION SCAN COMPLETE
===========================
File: [path or "direct input"]
Risks found: X
Auto-fixed: Y
Needs user input: Z
FIXES APPLIED:
- "SQL" → "S-Q-L" (line V1:3)
- "Linux" → "Lin-ucks" (line V2:5)
NEEDS USER INPUT:
- "live" (line C:1) - lyve or liv?
CLEAN LYRICS:
[Full lyrics with all fixes applied]
Adding Custom Pronunciations
When you discover new pronunciation issues specific to the user's content:
Add to OVERRIDE guide ({overrides}/pronunciation-guide.md):
- Read config to get
paths.overrideslocation - Check for
{overrides}/pronunciation-guide.md - Create file if it doesn't exist (with header and table structure)
- Add the word to appropriate section (Artist Terms, Album Names, etc.)
- Include: word, standard spelling, phonetic spelling, notes
Example entry:
| Larocca | larocca | Luh-rock-uh | Character in "sample-album" album |
DO NOT edit the base guide (${CLAUDE_PLUGIN_ROOT}/reference/suno/pronunciation-guide.md) - plugin updates will overwrite it.
When to add:
- Artist names, album titles, track titles
- Character names in documentary/narrative albums
- Location names specific to album content
- Any pronunciation discovered during production
This keeps discoveries version-controlled with the music content in the overrides directory.
Remember
- Load both guides at start - Base guide + override guide (if exists)
- Homographs are landmines - live, read, lead, wind WILL mispronounce without fixes
- Tech terms need phonetic spelling - Don't trust Suno with acronyms
- Non-English names always need help - Phonetic spelling mandatory
- Numbers are tricky - Write them out or use apostrophes
- When in doubt, ask - Better to clarify than regenerate
- Add discoveries to OVERRIDE guide - Never edit base guide (plugin will overwrite)
Dépôt GitHub
Compétences associées
release-standards
DocumentsCette compétence fournit des directives de gestion sémantique de version (semver) et des normes de formatage de journal des modifications pour les publications logicielles. Utilisez-la lors de la préparation des versions pour incrémenter correctement les numéros de version (majeure/mineure/corrective) et structurer les entrées du journal des modifications. Elle inclut des règles pour les identifiants de pré-version et des exemples clairs pour les développeurs.
commit-standards
DocumentsCette compétence formate les messages de commit Git selon la norme Conventional Commits. Elle fournit des modèles et des définitions de types (comme `feat`, `fix`, `refactor`) pour garantir la cohérence lors de la rédaction ou de la revue des commits. Utilisez-la pendant le processus de commit pour créer un historique de commits clair et structuré.
huggingface-tokenizers
DocumentsCette compétence offre une tokenisation haute performance en utilisant la bibliothèque basée sur Rust de HuggingFace, traitant 1 Go de texte en moins de 20 secondes. Elle prend en charge les algorithmes BPE, WordPiece et Unigram, tout en permettant l'entraînement de tokeniseurs personnalisés et le suivi des alignements. Utilisez-la lorsque vous avez besoin d'une tokenisation rapide pour la production ou pour créer des tokeniseurs personnalisés intégrés à l'écosystème transformers.
nano-pdf
Documentsnano-pdf est un outil CLI qui permet aux développeurs de modifier des PDF à l'aide d'instructions en langage naturel, comme changer du texte ou corriger des fautes de frappe sur des pages spécifiques. Il est idéal pour des modifications rapides et programmatiques de PDF directement depuis le terminal. Vérifiez toujours le résultat, car la numérotation des pages peut varier entre les versions.
