Don’t regenerate your song.
Save it.
Song Surgery is line-by-line lyric repair until your song is release-ready. Per-issue verdicts. Per-arc Studio modes. Verified Release Clearance.
The wrong question, and the right one.
“Should I regenerate this song?”
Throws away the lines you already loved. Burns credits. Wastes the work the war room already did. And the new generation has its own problems.
“Where, exactly, does this song need surgery?”
Save the lines that work. Replace only the lines flagged by the audit engines. Walk the triaged list one card at a time — Accept / Edit / Reject / Skip.
Restraint is the trust signal.
When the engines find nothing flagged — no QA gate failures, no taste flags, no line-level issues — Song Surgery says so. Not “here are some optional polish suggestions.” It says:
“This song doesn’t need surgery. Composite 92/100, Release Readiness 88/100. Ship it.”
A surgery that always finds something to operate on is a gimmick. We’d rather refuse to suggest a change than invent one.
Four actions. One verdict. Every line.
Each surgery item gets its own card: the current line, why it matters, a suggested revision. You choose what happens. The judge primitive scores the edit you ship.
Apply the suggested revision (or edit it first). The judge primitive scores the edit across 4 dimensions (specificity / singability / memorability / vulnerability) and surfaces an Improves / Changes / Risks verdict.
The suggested revision is a draft, not a mandate. Edit in place before accepting — every revision is yours, not the system's.
The suggestion is wrong. The line stays as it was. The surgery doesn't argue.
I'll come back to this. Different from Reject — you're not deciding the suggestion is bad, just deferring the choice.
Nine arcs. Nine surgeries. One discipline per genre.
Studio tier swaps the generic “Song Surgery” header for the arc-branded variant on supported genres. The byline names the load-bearing Sacred Accident the arc’s surgery enforces. Same engine; arc-specific craft signature.
- Worship SurgerySA#24
“The divine subject must be NAMED, not implied — Berean Test.”
Ambiguous pronoun ratio (APR), tradition-faithfulness index (TFI), divine-glory specificity, congregational singability (CSR), Berean vetting (BVT).
- Country SurgerySA#20
“Authenticity is INHABITED, not INHERITED — no cosplay verses.”
Concrete image density (CID), local rhyme rate (LRR), modal cadence (MSC), vernacular integrity (VIG), specific image instance (SII).
- R&B SurgerySA#23
“Vulnerability requires receipts — the AID rule, line by line.”
Narrative concrete density (NCD), hook vulnerability (HVT), action/image/detail per claim, melisma fitness, vocal-dialect authenticity.
- Pop SurgerySA#21
“Phonetic mass beats semantic precision — sing-along, first listen.”
Phonetic vowel rhyme (PVR), chorus hook rate (CHR), structural shape rotation (SSR), title-to-hook trajectory, AICR + FMR.
- Rock SurgerySA#27
“Open vowels at the chorus peak — the vowel must yield to the pitch.”
Vowel-pitch integrity (VPI), register-vowel landing (RVL), tension/release (TTI), rhythmic hook contour, gang-driving downbeat.
- Indie SurgerySA#25
“The personal detail is the universal door — hyper-specificity wins.”
Hyper-specificity index (HSI), slant-assonance rhyme (SAR), lyric-detail integrity (LDI), narrative-shape commitment, authenticity-posture test (APT).
- Folk SurgerySA#26
“Structure IS feeling — asymmetry is grief, symmetry is acceptance.”
Structural deviancy value (SDV), prosody-greedy-index (PGI), narrative-arc presence (NAP), rhythm-stress consistency (RSV), specificity-memory mirror (SMM).
- Latin SurgerySA#22
“A dialect is OWNED, not assembled — no Tourist Spanish.”
Dialect consistency score (DCS), dialect coherence test (DCT), Spanish phonology + ARI + CRI, tourist-Spanish guard, code-switching coherence.
- Rap SurgerySA#19
“A genre we cannot evaluate cannot be a genre we can serve.”
Per-bar diagnostics: forced rhyme, dead bars, slant-at-peak, internal-rhyme fatigue, wordplay-over-story, hook obscuration, borrowed-slang-without-earning.
Projected, verified, honest.
Every accepted edit is judged across four dimensions (specificity, singability, memorability, vulnerability). Those per-edit verdicts roll into a projected per-axis score. At session close, a verified rescore re-runs the eval engine on the rewritten lyrics.
When the verified rescore is meaningfully worse than the projection, Song Surgery says so:
“Projected +8 across dimensions; verified +2 composite. The chorus is tighter but the bridge held the song back. Review which axes the edits actually improved.”
Per-edit projections are signals, not promises. The verified rescore is the truth.
Try a different path.
The surgery items don’t change. Your choices do. Revert to v1.0.0, walk the same items differently, see how the per-axis outcomes compare. Every attempt stays in the session — side-by- side, no work lost.
v1.0.0 → v1.1.0 → v1.2.0 ... the version timeline IS the artifact. Surgery is the discipline of choosing well.
Save your song.
Song Surgery ships with the Professional tier. Free + Creator see a preview (2 items, no Final Vitals, no branching). Upgrade for the full workflow + per-arc Studio modes.