---
name: ai-music-lyrics
description: Create, rewrite, optimize, translate, or adapt lyrics for AI music generators and pop songwriting workflows. Use when the user asks for song lyrics, AI music lyrics, Suno/Udio-style lyric prompts, rhyme improvement, hook writing, verse/chorus structure, multilingual lyric adaptation, or polishing existing lyrics for singability, emotion, structure, and generator-ready formatting.
---
# AI Music Lyrics
Use this skill to create production-ready lyrics or improve existing lyrics for AI music tools and pop songwriting. Prioritize singability, emotional clarity, memorable hooks, clean structure, and generator-friendly formatting.
## Required References
Load only the references needed for the task:
- For any full lyric draft, read `references/song_structure.md`.
- Before adding structure tags, vocal directions, singer handoffs, instrumental breaks, rap/solo sections, or parenthetical effects, read `references/lyrics_symbol_rules.md`.
- For lyric polishing, rewriting, rhyme improvement, translation, or adaptation, read `references/lyrics_optimizing.md`.
## Workflow
1. Classify the task as `create`, `optimize`, `adapt/translate`, or `advise`.
2. Capture the essentials: language, genre/style, mood, theme/story, singer persona, target duration, whether the lyrics are for an AI music generator, and whether the user explicitly requests bracket-based performance or arrangement control.
3. If critical details are missing, infer reasonable defaults. Ask at most three short questions only when the result would otherwise be unusable.
4. Select the section sequence in this order: follow an explicit user-specified structure; otherwise choose the best form for the requested genre, emotional contour, narrative movement, and target length; only when no requirement or strong structural cue exists, use the common fallback from `references/song_structure.md`.
5. Draft or revise with clear section contrast where those sections are actually needed:
- verses develop images and story
- pre-chorus raises tension
- chorus delivers the hook and central emotion
- bridge changes perspective or intensity
6. Add local AI control tags only when the user explicitly asks to encode vocal, delivery, instrumental, or arrangement details with `[]`. When authorized, place each tag exactly where the requested change begins.
7. Run a final quality pass for meter, rhyme, natural pronunciation, repeated hook strength, emotional progression, and AI tag correctness.
## Output Rules
- When the user asks for lyrics or lyric optimization, output the final lyrics first.
- Keep explanations out of the output unless the user asks for reasoning, options, or design advice.
- Use section tags consistently, such as `[Intro]`, `[Verse 1]`, `[Pre-Chorus]`, `[Chorus]`, `[Bridge]`, and `[Outro]`.
- By default, use square brackets only for structural section tags. Do not add local performance or arrangement controls unless the user explicitly requests detailed `[]` control.
- Do not treat genre, mood, language, singer gender, or a duet concept alone as permission to insert tags such as `[soft female vocal]`, `[emotional]`, or `[layered harmonies]`.
- Never use `[Final Chorus]`. If a closing chorus is earned, label it `[Chorus]`; if another climax would weaken the song, omit it.
- Do not reuse one structure for every song. Outside the common fallback case, let the song type and emotional arc determine whether `[Pre-Chorus]`, `[Chorus]`, `[Bridge]`, `[Drop]`, or other sections appear and return.
- When detailed control is requested, keep AI performance directions short and inside square brackets. Do not overload every line with tags.
- Use parenthetical backing vocals sparingly for echoes, harmonies, call-and-response, or emphasis.
- Do not copy or closely imitate existing lyrics. If the user asks for a living artist's exact style, convert it into neutral musical traits.
- For Chinese lyrics, keep lines easy to sing, avoid overly literary compression, and favor strong vowel endings for hooks.
- For English lyrics, prefer natural stress patterns over forced rhyme.
## Default Assumptions
When the user gives no constraints, write:
- language: same language as the user's request
- length: short radio/AI song, about 2:30
- structure: `[Intro]`, `[Verse 1]`, `[Pre-Chorus]`, `[Chorus]`, `[Verse 2]`, `[Chorus]`, `[Bridge]`, `[Outro]` only when the user gives no structure requirement and the song type provides no stronger structural direction
- non-structural square-bracket controls: none unless explicitly requested
- tone: polished commercial pop unless another genre is implied
## Final Check
Before responding, verify:
- The song has one clear title or hook idea.
- The chorus is repeatable and stronger than the verses.
- Line lengths feel singable rather than essay-like.
- Rhymes support meaning instead of distorting it.
- AI tags follow `references/lyrics_symbol_rules.md`.
- When requested, vocal handoffs, instrumental breaks, and parenthetical harmonies appear at the intended musical moment.
- Every non-structural square-bracket tag can be traced to an explicit user request for detailed lyric control.
- The structure follows the priority of explicit user request, song-type fit, then common fallback.
- The common fallback, when used, follows `[Intro]` → `[Verse 1]` → `[Pre-Chorus]` → `[Chorus]` → `[Verse 2]` → `[Chorus]` → `[Bridge]` → `[Outro]` exactly.
- The output contains no `[Final Chorus]` tag.
