Suno · vocal timbre reference

Vocal Timbre Library

Vocal timbre reference for Suno — mapped across register, grit, and affect. Built for the latest Suno.

The library exists so we always know what vocal ground we’ve already covered — by genre and sub-genre. Every entry records an ungendered timbre we’ve used and the sub-genre it came through, which frees us to keep reaching for something genuinely different each time instead of unknowingly repeating a voice, a genre, or a sub-genre we’ve already done. The genre and sub-genre are the vehicle; the real target is a distinct vocal timbre, and tracking is what lets us spread those voices deliberately across registers, grit, affect, and an ever-widening range of genres and sub-genres rather than clustering. In short: it’s a map of explored territory — voices and sub-genres alike — that keeps pushing us into unexplored ground.

Exploration, not prohibition: repeats of genre, sub-genre, or timbre are acceptable and expected — the same prompt can yield different results, and popular ground gets revisited on request. Tracking informs; it doesn’t block.

Prompt architecture

  • Descriptor-only — no artist or band names, ever.
  • Positive descriptors in the style block; all negations in the separate negative field (no inline negation).
  • Style cap 1000 chars · lyric cap 5000 chars.
  • Bracketed lyric text = production instruction, not sung.
  • Anchor phrase: a unique and specific vocalist (batch 4+; earlier batches used "a single and specific vocalist").
  • Suno defaults to guard against: orchestral wash, wrong vocal gender, blended choir — negatives are essential.

Inspirations mode (artist-anchored prompts)

  • To mirror an artist's style, describe their inspirations, not their sound: reverse-engineer the artist into ancestral ingredients (genres, eras, gear, scenes, constraints) and prompt the recipe. Sources carry deeper training signal than "sounds-like" descriptions; still descriptor-only, no names anywhere.
  • Standard set per artist: one blend prompt (all roots married) plus isolated root prompts, one per inspiration lane. Roots can be mixed pairwise to triangulate specific songs.
  • Temperament rule: shared ancestry is not enough — encode the attitude the artist brought to the ancestors (menace, restraint, mania) or the prompt yields their politest cousin.
  • Chassis-swap technique: any root prompt can be transplanted into another genre frame (e.g. a piano-soul root into space techno) by keeping the temperament and vocal character while re-rendering instrumentation in the new frame's vocabulary; texture dialects (dub, Detroit strings, microsound, modular drift, progressive layering) are swappable one at a time against a fixed skeleton.
  • Inspirations studies are promoted into the library as cat:'inspiration' entries (searchable by hidden artist key): garage blues-rock duo #147–157, mid-90s spiritual alt-rock #158–163, mid-70s California rock #164–169 (+ space-techno chassis swaps, doc appendix), machine-gospel darkwave #137–141, ritual prog-metal catharsis #142–146. The companion artist_studies.md is generated from these by tools/build.mjs.

Reese & low end

  • Reese = single transient at structural breaks only, non-rhythmic, non-repeating.
  • "Climax" and "story" trigger unwanted repeating structures — avoid.

Eight mandated negatives (every entry)

  • Harmonic stagnation · Dynamic flattening · Quantized sterility · Structural monotony · Predictability · Vocal identity drift · Vocalist substitution · Soundtubes
  • Note: "Quantized sterility" fights gridded electronic genres — aimed at lifelessness, not the beat; first term to pull if the pocket loosens. Pull it outright on #367–371, where the pitch machine is the instrument, and on #428, where the grid and the pitch correction are both the subject.
  • "Structural monotony" and "Predictability" fight anything built on deliberate non-development. Pull both on #427, #429 and #434, where nothing changing for minutes at a time is the piece — a derelict-ship drone, a motorik pulse that never varies, and a drift built on held pads whose only event is the doubling cancelling itself. On #434 pull "Harmonic stagnation" as well: one struck note every eight bars is the ask, not a defect. On #428 pull "Dynamic flattening" too: that entry asks for no expression whatsoever.
  • "Vocal identity drift" and "Vocalist substitution" guard voice continuity: "drift" fights the lead's timbre morphing across the track, "substitution" fights a different singer appearing in a later section. Pull both on wordless / choir-as-instrument entries, where the voice is meant to change.

Eight compositional terms (woven where genre supports)

  • Melodicism · Voice-leading · Tension-release · Dynamic contrast · Harmonic movement · Push and pull · Hold release · Dissonance and resolve
  • Withheld from monophonic / harmonically static traditions (modal chant, maqam, drone).

Nine-stage pipeline (for full songs)

  • Idea → Tone → Symbolism → Structure → Prosody (draft) → Meter → Flair/Vocalese → Cathartic-to-sing → Suno-fy. Run sequentially, no mid-pipeline pauses.

Register × grit map · every entry, read from its own prompt

cool / detached melancholic / weary earnest / yearning swagger / playful ecstatic / fervent anguished / torn menacing / aggressive
hollow = the prompt doesn't name that axis (drawn neutral)

Prompt library

Electronic shelf · 81–95, 106–120, 131–136, 175–179 (role · energy · processing)

#VoiceSubgenreBPMVocal roleDefining treatment