Open Standards
Three open standards plus our public Sacred Accidents log — the four artifacts that document the disciplines this product operates under. The trio is licensed CC BY 4.0 (fork them, cite them, audit yourself against them); the log is our running register of the truths the WAR Rooms surfaced.
Lyric Scoring Standard v1.2
The 12-metric rubric SongForgeAI scores every lyric against.
Twelve craft metrics across Craft (25%), Expression (40%), and Impact (35%) — with anti-inflation rules (Gravity 50, Burden of Proof, Antagonist Ceiling, Historical Context). Published spec; published whitepaper; published per-metric deep pages; signed-verdict ed25519 envelopes for verifiable scoring. The category-noun claim.
Read the standardBanned Clichés List · 327 terms
The terms we filter from every lyric, by enforcement tier.
Three tiers — hard-banned (always rewritten), defense-required (must justify in writing-room notes), watch-list (overuse tracked). 327 terms across AI-cliché, overworn-imagery, filler, and purple-prose categories. Plus vertical-specific extensions for memorial ("loved one", "earned their wings") and wedding ("soulmate", "happily ever after").
See the listVoice-Reference Discipline · 58-voice rosters
How we use named artists as craft references without imitating them or naming them in delivered output.
Five-layer enforcement: forge prompt → refine prompt → cold reader → roster-name scanner (with 9-test coverage) → auto-rewrite of any leak. The MFA-syllabus pattern, mechanically verified. Sacred Accident #15.
Read the disciplineSacred Accidents · 9 of 19 documented
The truths the WAR Rooms surfaced while doing other work. Public register of what this product will and will not do.
The Sacred Accidents are not a forkable standard — they're a brand posture made visible. SA#11: it is more comfortable to debate strategy than to send a cold DM. SA#12: the product can do everything except be there. SA#13: the song must survive being heard once. SA#14: specificity is not optional. SA#15: inspiration is reference; imitation is theft. SA#16: the system writes 100 variations on the same three internal monologues — the cure is cross-song memory. Six named disciplines; ten more awaiting reconstruction.
Read the logWhy publish these
Trust transparency. Anyone buying a $99 heirloom from us — or evaluating us for partnership, journalism, or counsel — can read the discipline before signing on. The public claim and the internal practice match.
Category leadership. The AI music space has dozens of generators and few published standards. Naming the disciplines we operate under, and licensing them so others can adopt, is the standards-leadership move every category-defining product makes.
Ratchet against drift. Each artifact is the contract the codebase ratchets against. CI checks verify the artifacts cross-link (since B2675) + that the banned-cliché list stays dynamic + that the voice rosters never leak into delivered output.
CC BY 4.0 — fork these
Use any of these in your own tool. Attribution appreciated.
If you build a lyric tool or a creative-AI product and you fork the scoring rubric, the banned-cliché list, or the voice-reference discipline, a link back to songforgeai.com is appreciated. The artifacts are versioned; the spec evolves transparently; the audit trail is public.
Citation guidance lives on each individual standard page. The full source for each artifact is in this codebase under src/lib/scoring/, src/lib/banned-terms.ts, and src/lib/life-songs/voice-rosters/.