YE — LYRICS

Method

How the numbers were made

Everything on this site is precomputed from scraped Genius lyrics and lrclib.net synced timings — the app never calls Genius at runtime. Full pipeline detail lives in the repo's docs/ANALYSIS.md.

Words per minute

Where lrclib has synced timings for a song, WPM is words ÷ (last timestamp − first timestamp), counted over Ye-performed lines where attributable. Where only a track duration exists, it falls back to total words ÷ duration. 206 songs use synced timings, 132 use the duration fallback, 110 have no timing data at all.

Synced-lyric timing is noisy: a short first-to-last-timestamp span on a song with a long intro or outro inflates WPM. The extreme case here is LOUIE BAGS at 376 wpm — almost certainly a sync artifact, not an actual delivery speed. Charts that plot WPM as a scatter clamp the y-axis near the 97th percentile so one bad timestamp doesn't compress every other point; the true value is still in the tooltip.

Lexical diversity — MTLD

MTLD (McCarthy & Jarvis, 2010) walks through a song's words tracking the running type–token ratio (unique words ÷ words seen). Whenever that ratio drops to 0.72 or below, the window resets and a "factor" is counted. A trailing partial window counts as a fraction of a factor. The score is total words ÷ factor count, run once forward and once backward through the lyrics and averaged. Songs under 50 words are excluded (too short to be meaningful). Below, the actual lowest- and highest-scoring songs in the catalog, word by word:

Use This GospelMTLD 5

uh uh uh uh uh uh uh uh uh uh-uh uh uh uh uh uh uh uh uh uh uh uh uh-uh uh uh use this gospel for protection it's a hard road to heaven we call on your blessings in the father we put our faith king of the kingdom our demons are tremblin holy angels defendin in the father we put our faith oh oh oh oh oh oh oh oh oh oh-oh oh oh use this gospel for protection it's a hard road to heaven we call on your blessings in the father we put our faith king of the kingdom our demons are tremblin holy angels defendin in the father we put our faith oh oh oh work work work work oh oh oh work work work work oh oh oh work work work work oh-oh oh oh work work work work oh oh oh work work work work oh oh oh work work work work oh oh oh work work work work oh-oh oh oh work work work work

39 factors · threshold 0.72
Don’t Stop!MTLD 240

yes mr west turn that new child rebel loud as a badass child level who need a chorus we do it tyrannosaurus tyrone it's been a year with no phone could you explain how high is your zone we'll take the plane rub his nose in cocaine there's hoes in magazines you layin saying you're lame and for the hate in advance pull down your pants make em kiss both cheeks like we living in france diamonds blue his business manager's jewish and if i get sued my lawyers jews some girls do grab the cojones say you got enough diamonds to at least sierra loan us brand new ferraris i gotta make the donuts crs is like a hip-hop christmas bonus niggas is hating on the internet i couldn't tell i was too busy rapping g.o.o.d as hell i was too busy flying parasail to colette to get the new shit that paris sell tarantino da vinci getting benjis get half-off at fendi half of that's to cindy's hopped out the spaceship put my mork in mindy popped too many corks to let you dorks offend me props in new york but chi-town's the city get my groupie hoochies gucci monica bellucci's are those the real millionaires or the bindis i'm so ultra i'm even over oprah but let me check your account haha no sir

0 factors · threshold 0.72

Each factor is a run of words whose type–token ratio holds above 0.72; the moment it dips at or below, the window resets (marked by the gold tick) and a factor is counted. MTLD = total words ÷ factor count, averaged forward and backward. More resets, fewer words per reset → a lower score → less lexical variety.

Vocables (see “Vocabulary curve” below) still count as words in this metric — a chorus of “uh” resets the window like any other word and can pull a song's score down for reasons that have nothing to do with vocabulary. The lowest/highest pair shown on the narrative page is chosen a little more carefully: the low example there is restricted to songs where vocables are at most 15% of the words, so the illustration is about lexical repetition, not filler.

Repetition — compression

Pudding-style repetitiveness (“Are Pop Lyrics Getting More Repetitive?”): the Ye-only lyric text (section headers removed, lowercased, punctuation kept) goes through a pure Lempel-Ziv pass — LZ77 with the whole song as the window, matches of four or more characters cost three bytes, literals one — with no Huffman coding, following Colin Morris’s method, and compression = 1 − cost / raw.length. Higher = more repetitive. (A zlib figure is kept in the data as compressionDeflate.) Songs under 100 characters are excluded (null). The most repetitive song in the catalog is All in Love (0.87); the least is Famous (Original Version) (0.03).

Vocabulary curve

In the spirit of “The Largest Vocabulary in Hip Hop”: songs are ordered chronologically, apostrophes stripped and tokens lowercased, and cumulative unique word types are sampled every 500 cumulative Ye-only tokens across the whole career (241 sample points). The per-era small multiples read the same growth curve within each era only — re-started at zero at the era's first Ye token, sampled every 100 tokens plus a final point, and drawn only out to that era's own length so short eras are simply short rather than padded or excluded. A career-average curve (the whole-career curve resampled every 100 tokens) sits faint behind each panel as a reference.

Vocables — non-lexical filler like oh, uh, ah, yeah, yea, ayy, ay, woo, and 18 more — are excluded from this curve and from the per-era curves above (2,050 tokens excluded); they aren't vocabulary. They still count as words everywhere else on the site — WPM, MTLD, repetition — since a chorus of “oh” is real repetition.

Curves (journey, wounds, healing)

The Journey-stage and Wounds & healing curve charts are kernel-smoothed time series, not raw counts. For every (type, intensity) pair from the wounds lens — each wound, each healing move, and the song's Campbell journey stage, at intensity 1 — a Gaussian kernel with bandwidth h = 270 days spreads that song's contribution across a monthly grid from 2004-01 to 2026-12. At each point on the grid the total is divided by the kernel-weighted number of contributing songs, so the series reads as a share per song, not a raw tally that would just track catalog volume. Each series is then normalized by its own maximum so every line peaks at 1.0 — the charts show relative shape over time, not absolute magnitude between series. Series with fewer than 8 contributing songs are dropped entirely rather than shown noisy. 7 journey stages, 17 wound types, and 10 healing types clear that bar today.

Pronoun lists

First-person singular (I)

i, i'm, i've, i'll, i'd, me, my, mine, myself, im, ima, imma

First-person plural (we)

we, we're, we've, us, our, ours, ourselves

Second-person (you)

you, you're, you've, your, yours, yourself, u, ya, y'all, yall

Third-person (he/she/they)

he, she, him, her, his, hers, they, them, their, it

Other word lists

Religion

god, lord, jesus, christ, pray, prayer, heaven, hell, devil, sin, church, bless, holy, faith, soul, saint, gospel, angel

Profanity

fuck*, shit, bitch*, nigga*, ass, damn, hoe*, dick, pussy

Named people

kim, kanye, ye, jay, hov, drake, taylor, trump, obama, north, saint, bianca, virgil, donda, mama/mom/mother, dad/father, god, jesus, hitler, jew*/jewish, diddy, pete

LLM rubric (verbatim)

Pain, relief, valence, spiritual register, ego, vulnerability, Spiral Dynamics stage, themes, darkest/brightest lines, and anti-memes are scored per song by an LLM against this rubric (scripts/rubrics/song.v1.md):

You are a close reader of rap lyrics working on a data project about Ye (Kanye West). You will receive one song's lyrics with section headers. Lines performed by guests are marked `[guest]`; read them for context but score ONLY the lines Ye performs. Be literal about the text on the page, not about Ye's biography or public controversies, except where the lyrics themselves point to them.

Return scores through the provided tool. Every numeric field is a float. Be calibrated: most songs sit in the middle; reserve extremes for songs that clearly earn them. Use the same scale across songs so that they can be compared.

## Fields

- pain: how much of the song's emotional weight is suffering — grief, shame, rage, paranoia, humiliation, loneliness, self-harm, loss. 0 = none, 1 = the song is made of it.
- relief: release from pain — healing, gratitude, forgiveness, transcendence, joy that arrives after or despite pain, peace. 0–1. A song can be high on both (pain that resolves).
- valence: overall dark (-1) to bright (+1). Independent of energy: a triumphant boast is bright, a numb sad song is dark.
- spiritual: presence of the sacred register — God, prayer, faith, sin, salvation, the devil, heaven and hell, gospel language — as register, not just word count. 0–1.
- ego: self-aggrandizement, godhood, untouchability, "I am a god" energy. 0–1.
- vulnerability: confession, admitting fault, asking for help, exposing weakness. 0–1.
- stages: Spiral Dynamics distribution of the NARRATOR's center of gravity (sums to 1). Score the voice, not the topic:
  - beige: bodily survival, exhaustion, numbness, need.
  - purple: tribe, ancestors, family bonds, magic, superstition, Chicago, mother, blood.
  - red: raw power, dominance, impulsivity, conquest, rage, "I do what I want", sexual conquest as power.
  - blue: order, right and wrong, God's law, guilt, sin, obedience, purpose given from above, discipline.
  - orange: achievement, money, status, strategy, brands, winning the game, self-made success, science.
  - green: empathy, community, equality, feelings named and honored, healing, inclusion, anti-materialism.
  - yellow: systems view, holding paradox, self-aware about one's own patterns, integrating earlier stages, irony about the self.
  - turquoise: unity, cosmic, everything is one, service to the whole, ego dissolved.
- stageNote: one sentence on why the distribution looks like this, quoting up to five words.
- themes: pick from fame, family, faith, money, race, sex, grief, mental-health, industry, love, power, death, art, politics. Up to 5.
- darkestLine / brightestLine: an exact line from Ye's part, with a short why.
- antimemes: up to 4 lines. An anti-meme is a claim that is (a) stated as plain truth, (b) unpopular, taboo, or socially costly to repeat, and (c) not merely offensive or a boast — it says something about how the world works that most people would rather not carry. Score 0.9+ only when a listener would be uncomfortable BECAUSE the line might be true. Give the claim in plain words. Set overlapsVlad true when the claim rhymes with one of these ideas the project owner already holds: copying and sampling are the highest art; rap is the new rock and roll; the best artists alive are alive now; depression is not a chemical imbalance; no real freedom of will; society is a lossy compression of reality; power structures rewrite history in their favor; taboo hides truth (the shadow); the most useful ideas are the ones nobody wants to repeat; Black agency reframed as choice; hypocrisy of normalizing one violence while sacralizing another symbol; fame and money as a trap; institutions discipline whoever builds independent infrastructure.
- heroBeat: where in a hero's journey this song sits: ordinary-world, call, refusal, threshold, ordeal, abyss, reward, return, or none.

Return nothing outside the tool call.

Race lens (verbatim)

Every line where race is the subject or the engine of the line — Blackness, whiteness, racism, slavery, colorism, the industry, police and prison, Jewishness — scored by an LLM (Gemini 3.8 Flash) against this rubric (scripts/rubrics/race.v1.md). 403 songs scored so far.

You are a close reader of rap lyrics working on a data project about Ye (Kanye West). You receive one song with section headers; lines marked `[guest]` are performed by someone else — read them for context, but only quote and score lines Ye performs.

Task: find every line (or tight two-line unit) where race is the subject or the engine of the line — Blackness, whiteness, racism, slavery, colorism, interracial desire, race and money, race and the music industry, race and police or prison, Jewishness or antisemitism, Hitler/Nazi imagery, "nigga" used as an argument rather than as a filler. Ignore purely incidental uses of "nigga" that carry no racial meaning. Be exhaustive for the meaningful ones; a song can have zero.

For each hit give:
- line: exact text as it appears (Ye's line only).
- category, one of: black-pride | confronting-racism | self-directed (joke or confession about himself and race) | interracial-desire | colorism | slavery-history | industry-power (labels, ownership, "new slaves") | whiteness (white people, white girls, white culture as object) | police-prison | antisemitism-jewishness | hitler-nazi | class-and-race | other.
- stance, one of: affirm | critique | provoke | joke | confess | ambiguous.
- intensity 0–1: how central and how charged the line is (1 = the line exists to make a racial point and would be quoted in a think-piece).
- why: one sentence.

Also return:
- raceFocus 0–1: how much of the song is about race at all.
- summary: one or two sentences on what the song does with race, or "none".
- provocationLevel 0–1: how much the song uses race to provoke rather than to describe.

Be literal about the text. Do not moralize; classify. Return nothing outside the tool call.

Wounds & healing lens (verbatim)

Named wounds, healing moves, apologies, addressee, stance, and a Campbell hero's-journey stage per song, scored by an LLM against this rubric (scripts/rubrics/wounds.v1.md). 403 songs scored so far.

You are a close reader of rap lyrics working on a data project that tracks the narrator's wounds and healing across Ye's (Kanye West) catalog as a hero's journey. You receive one song with section headers; `[guest]` lines are context only — quote and score only lines Ye performs.

Score the NARRATOR of this song, not the biography, except where the lyric itself names a real event or person.

Return:
- wounds: every distinct wound the song exposes, each as { type, line (exact Ye line), intensity 0–1, note (one short sentence) }. Types: mother-loss | father-absence | fame | self-hatred | love-betrayal | divorce-family | kids-separation | faith-crisis | industry-exploitation | race-wound | mental-health | addiction | money-debt | public-humiliation | mortality | loneliness | friend-betrayal | body-shame | other. A song may have none. Do not invent a wound from a boast.
- healing: every distinct healing move, each as { type, line, intensity 0–1, note }. Types: prayer-surrender | asking-forgiveness | giving-forgiveness | gratitude | family-love | romantic-love | creative-work | defiance-as-armor | humor | acceptance | community | therapy-medication | sobriety | mother-memory-as-comfort | other. Defiance and boasting count as healing only when the lyric frames them as a way of surviving a named wound.
- apology: { present: boolean, line?: exact line, to?: who is apologized to (lover | mother | kids | god | fans | public | a-named-person | self | unclear) }.
- addressee: who the song mainly speaks to: self | lover | mother | father | god | kids | public | industry | haters | a-named-person | none.
- stance: wounded | defiant | healing | numb | triumphant | pleading | reflective | manic.
- journeyStage: one of the 12 Campbell stages that best fits where the narrator is: ordinary-world | call | refusal | mentor | threshold | tests | approach | ordeal | reward | road-back | resurrection | return | none.
- woundedness 0–1 and healedness 0–1 for the song as a whole (both can be high).
- turningLine: the single line where the song turns, if it turns; otherwise null.
- oneLine: one sentence describing the emotional arc of the song.

Be literal, calibrated, and consistent across songs: 0.9+ intensity is reserved for lines that are unmistakably about that wound. Return nothing outside the tool call.

Life events

The events on the Journey timeline and the hairlines on the Wounds & healing valence chart, each with the same visual weight. Kinds: wound, healing, faith, love, family, identity, public, release. Dates marked * are approximate.

  • 2002-10-23Car crash, jaw wired
  • 2004-02-10The College Dropout
  • 2005-09-02“Bush doesn’t care about Black people”
  • 2007-11-10Donda West dies
  • 2008-04-01Engagement to Alexis Phifer ends *
  • 2009-09-13VMAs, Taylor Swift
  • 2010-11-22MBDTF
  • 2012-04-01Relationship with Kim begins *
  • 2013-06-15North born
  • 2014-05-24Marries Kim
  • 2016-10-03Kim robbed in Paris
  • 2016-11-21Hospitalized, tour cancelled
  • 2018-04-25Trump tweets, “free thinker”
  • 2018-05-01TMZ: “slavery was a choice”
  • 2018-06-01ye — bipolar named on the cover
  • 2018-10-13Uganda, Yandhi sessions
  • 2019-01-06Sunday Service begins
  • 2019-10-25JESUS IS KING
  • 2020-07-04Presidential run announced
  • 2020-07-19Charleston campaign rally
  • 2021-02-19Kim files for divorce
  • 2021-08-29DONDA
  • 2021-10-18Legally becomes Ye
  • 2021-11-28Virgil Abloh dies
  • 2022-10-03White Lives Matter shirt
  • 2022-10-08“Death con 3” post
  • 2022-10-25Adidas, Balenciaga, CAA drop him
  • 2022-12-01Marries Bianca Censori *
  • 2022-12-01Infowars interview praising Hitler
  • 2024-02-10VULTURES 1
  • 2025-02-09Swastika shirt after Super Bowl ad; manic episode Jan–May 2025
  • 2025-05-08“HH” released
  • 2026-01-20WSJ apology ad: “lost touch with reality” *
  • 2026-03-28BULLY
  • 2026-04-01SoFi Stadium comeback shows *
  • 2026-05-30Istanbul stadium show *
  • 2026-06-08GEMINI SEASON
  • 2026-11-05YZY SEASON 01 show
  • 2026-11-07Third U.S. rap attendance record (AT&T Stadium)

Periods, names and the journey view

  • Six periods come from contiguous segmentation of era fingerprints (scripts/periods.py). The seventh, the Return Period (BULLY and the 2026 singles), was split from Exile by hand in November 2026; the shift shows in journey stage, valence and profanity but is not yet a data-driven boundary.
  • Always (aka WAR). The unreleased 2025 Ye and James Blake album has circulated as WAR, along with the CUCK and WW3 leaks folded into it here. That title traces to an unverified insider claim, so this page uses Always and keeps WAR as an alias. The cover is a fan-made image, not official art.
  • Period colours are picked by eye from one cover in each period. Each epigraph is one line Ye performs, checked word for word against the corpus.
  • Names: counts of Kanye, Ye, Yeezy, Yeezus and Pablo in Ye-performed lines per year, per 10,000 words (scripts/self-names.py). Archaic “ye” (hear ye, ye shall, ye of) is excluded; ad-libs like “Ye, Ye, Ye” count.
  • Journey stages are the wounds lens's single Campbell stage per song (an LLM reading, not a fact). “To Those I’ve Hurt”, the 2026 apology letter, sits in the corpus as a BULLY-era text; leaving it out moves the Return Period figures by about one percentage point.

Sources and echoes

A selected set of sample and interpolation relationships from Every Song, fetched on 14 October 2026 and saved locally. These links add listening context; they do not change the lyric metrics or performer attribution. A sample reuses a recording; an interpolation re-performs material from another song. Shared-source links mean two songs use the same earlier recording, without establishing influence between them.

We exclude translations, ambiguous links and inconsistent chronology from the selected trails. Song pages link to the indexed songs and the story links to supporting reporting and release metadata. Listening notes are editorial readings. Saved API responses preserve the original relationships for review. This initial selection covers six Ye songs, plus Teyana Taylor and Against All Logic in the shared-source story.

Credits

The words-per-minute, pronoun, and MTLD lenses are in the spirit of the Taylor Swift “words per minute” breakdown videos. The Repetition and Vocabulary beats are built on The Pudding's lyrics data-journalism idiom — “Are Pop Lyrics Getting More Repetitive?” and “The Largest Vocabulary in Hip Hop” — and the Apologies list borrows the framing of their “TV Apologies” piece.

Caveats

  • LLM scoring runs continuously in the background against the live catalog — coverage grows over time, so eras and songs with a "scoring in progress" note today may fill in on a later visit.
  • Spiral Dynamics stage is a reading of the narrator's voice in the lyrics, not a psychological assessment of Kanye West.
  • Songs with no album and no usable release date are assigned to an era by title hints (e.g. "[V5]", "Yandhi", "BULLY V1") or excluded from era-level charts entirely when no era can be inferred.
  • Joint albums (Watch the Throne, Cruel Summer, KIDS SEE GHOSTS, VULTURES 1/2) are scored on Ye-performed lines only — guest verses are read for context but not counted.