What a 60 Means in the Lyric Scoring Standard

The first anti-inflation rule in the published Lyric Scoring Standard is the Gravity Rule: every metric defaults to 50. Not 70 with bonus points for trying. Not 80 unless something is "very wrong." 50 is the median; you have to earn every point above it.

Why 50 and not, say, 75? Because 75-as-baseline is what makes other AI tools’ scores meaningless. If the easy region of the score is 70-95 and the hard region is 95-100, the score communicates almost nothing — every output looks "pretty good." A 78 in that scale is indistinguishable from a 72 from any practical standpoint.

50-as-baseline reverses this. Most AI lyric output lands between 50-65 because that’s actually where output sits when measured against professional craft. A 78 then means real differentiation — six bands above the average. A 90 is genuinely rare. Anyone calibrating to the published 21-song corpus sees the math: hand-scored S-band entries (Hank Williams, Joni Mitchell, Marvin Gaye) sit at 92-95. AI output that hits 87 is in the company of canonical work.

The published grade scale + the meaning of each band:

• S+ (96+). Vanishingly rare. The published corpus has no entries this high; the rubric’s S-band anchor is 95. A 96+ has not been observed in routine evaluation. Treat 96+ output with skepticism — verify the seal, re-run, cross-reference.
• S (91-95). Canonical-level work. The published S-band anchors live here: Hank Williams, Joni Mitchell, Marvin Gaye, Caetano Veloso. Genuine S-band lyrics survive on artistic merit alone — they don’t need production to land.
• A+ (86-90). Excellent professional output. A working songwriter on a good day. AI output rarely reaches this band without targeted iteration; when it does, the lyric typically has at least one transcendent line + named-anchor specificity throughout.
• A (80-85). Strong professional work. Most "great" output falls here. Differs from A+ by missing one of the rubric’s S-band requirements (a transcendent line, a memorable hook return, etc.) but landing solidly elsewhere.
• B+ (73-79). Good professional output. Releaseable, won’t embarrass anyone. Most well-iterated AI output lands here. Lacks a "moment" but has no major weaknesses.
• B (65-72). Competent but generic. Recognizable as a song; doesn’t differentiate itself. Most first-draft AI output sits in this band.
• C+ (55-64). Above amateur, below professional. Has clear weaknesses + clear strengths. Often the band of partial drafts that need a verse rewrite.
• C (45-54). Median amateur. The rubric’s gravity floor — 50 is the metric default, so a composite of 50 lands here. Many "fine" lyrics live in this band; nothing wrong, nothing memorable.
• D+ / D / F (below 45). Real problems. F is reserved for output that fails on multiple metrics simultaneously. Most output never reaches this band; when it does, the gauntlet typically catches it before the user sees it.

If you’ve scored the same lyric on multiple AI tools, you’ve seen the spread: tool A says 87, tool B says 92, the SongForgeAI rubric says 71. Same lyric. Three numbers.

Three reasons:

1. Different scales. Most AI tools scale 70-100; SongForgeAI scales 0-100 with 50-as-default. Their 87 is roughly equivalent to our 67 in real-band terms.
2. Different rubric explicitness. Most tools score against an unpublished rubric; we score against the published one (CC BY 4.0). When you can’t see what they’re measuring, the score isn’t arguable. Our 71 is arguable: you can read which metric we marked you down on and disagree with that specific deduction.
3. Different anti-inflation discipline. The Anti-Platitude rule, the Burden of Proof rule, the Antagonist Ceiling — these are unique to the published Standard. They specifically prevent inflation that other rubrics tolerate.

This isn’t saying we’re right and they’re wrong; it’s saying our number means something specific that you can verify. Cross-checking against the published corpus shows the math directly: hand-scored canonical work sits where it sits, and your output relates to it numerically.

Practical: when the rubric returns a band, here’s what to do next.

• S-band (91+): Stop iterating. The rubric reads this as professional-craft-level work. Render it; release it; move on.
• A-band (80-90): One targeted pass. Identify the lowest single metric, rewrite the lines that earned that low band, re-score. A-band lyrics typically have one weak metric pulling the composite down; fixing it lifts the composite 3-5 points.
• B-band (65-79): Iterate or refine. The rubric’s automated gauntlet (every paid forge runs this) typically lifts B-band by 5-8 points on the rewrite pass. Manual Refine Mode with line-locking is faster than re-forging when you want to preserve specific lines.
• C-band (45-64): Restart or restructure. The rubric is telling you the lyric isn’t working at the line-by-line level. Either restart with a sharper concept or rewrite the structure (verse/chorus arrangement, narrator, time setting) before iterating on individual lines.
• D-band or F (below 45): Discard. The lyric has compound failures across metrics; targeted rewrites won’t lift it. Save the concept, restart the draft.

The rubric isn’t about hitting a high number for its own sake; it’s about which lines need work and why. A 71 with the per-metric breakdown ("M5 Specificity 58, M11 Memorability 62, M9 Transcendence 55") tells you exactly what to fix. A 87 with no breakdown tells you nothing.

Use the band as a quick read on whether the lyric is iterable from where you are. Use the per-metric breakdown to know what to iterate.