catalog

All songs, with the prompt and it’s the measurement of each one

Seven songs generated in Suno via Kie. For each one: what it is, the exact style prompt that was sent, the numbers measured in the resulting audio, and a player to listen. The idea is to view them side by side and understand why one sounded good and the other didn't.

All songs — summary

All 15 songs, new and from the collection, with the numbers for track 1 of each. Dynamics is the difference between the quietest and loudest part — the greater it is, the more the music breathes. Sustained notes counts the notes held for more than 0.35s. Hover over the card metrics to see what each one means.

music	bpm	key	dynamics	sustained	vibrato	verdict
Legacy (variant)	118bpm	F# minor	20.7 dB	29	112 cents	draft
Se Paga (PT)	118bpm	D# minor	16.3 dB	5	26 cents	constant intensity
It's Paid (EN)	118bpm	D# minor	14.8 dB	5	82 cents	constant intensity
Don't Give Up	76bpm	E minor	18.1 dB	13	70 cents	corrected contrast
The Center	84bpm	The smallest	24.7 dB	28	31 cents	corrected contrast
The Wheel	92bpm	C major	29.6 dB	12	68 cents	corrected contrast
Luz de Volta	82bpm	E minor	15.5 dB	3	—	new
Tears Become Light	—	A# minor	24.1 dB	16	17 cents	collection
We Were Here	—	C major	24.4 dB	44	37 cents	collection
Fire	—	G minor	22.1 dB	16	198 cents	collection
Horizons	—	C major	17.3 dB	11	66 cents	collection
Digital Sky	—	C major	17.1 dB	5	47 cents	collection
Builders	—	F major	13.7 dB	2	43 cents	collection
Isolated cover	—	G major	14.8 dB	0	—	collection
Cover reel	—	A# major	28.2 dB	2	—	collection

Catalog

Each generation returns two Suno tracks with the same prompt. They usually sound quite different from each other—listen to both before deciding.

The six charts for each track: o spectrogram is the sound map — time horizontally, frequency vertically (low at the bottom, high at the top), and color by volume; this is where the voice’s harmonics, vibrato, and melody lines appear. A waveform shows only the volume, useful for finding the chorus and verse. The piano roll translates the singing into blocks at the pitch of each key, showing the exact note. The movement is the pitch line with the volume behind it. The dialogue shows the phrases and the silence between the voices. And the histogram shows which notes the voice uses.

How to read the motion graph: the amber line shows the pitch of the voice throughout the song — when it goes up on the chart, the note goes up. Gaps in the line mark places where there is no singing (instrumental, breaths). The thin gray line behind it shows the volume. The notes at the left ends mark the lowest and highest notes the voice reaches, and the range in semitones tells you how big the jump between them is. A line that rises and falls is music that moves; flattened line at the top is music that lives in the shout and wears you out.

Legacy (variant)

The first draft of the theme: the price of success and inheritance. More restrained version, with 'emotional build' in the prompt.

draft

style prompt sent to Suno

epic cinematic pop, powerful female belt vocals, male spoken word responses, gang vocal choir, brazilian portuguese, orchestral war drums, soaring strings, anthemic, wide reverb, 118bpm, emotional build, huge chorus

negatives

lo-fi, mumble, whisper-only, acoustic ballad, autotune heavy

V5 model · style mode · female voice · styleWeight 0.85

clip history · clipe-v2 library · dark · slow cuts · almost no text

Almost the entire duration is in the story layer, with the stage appearing only briefly. Slow cut and just two cards: it's the contemplative version of the same theme.

prompts used in the images and animation

finish applied to EVERY image in the project

raw documentary photograph, live-action cinema still, anamorphic 35mm, natural skin texture and pores, film grain, practical light, shallow depth of field, muted teal and amber grade

negative, also across all of them

cartoon, 3d render, cgi, pixar, anime, illustration, painting, plastic skin, doll face, oversaturated, text, watermark, extra fingers, deformed

animation prompt template (Agnes)

Smooth cinematic transition between the keyframes: <movimento da cena>. Natural motion, consistent characters and style, cinematic camera.

The 22 scene descriptions and the 22 camera movements in this library are at v2-cenas.json e v2-movs.json. The complete step-by-step guide is at docs/clipes.

Track 13:01

20.7 dBdynamics

29sustained

112 centsvibrato

2685 Hzbrightness

vocal movement · key F# minor (confidence 0.79)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

Track 22:52

23.3 dBdynamics

32sustained

94 centsvibrato

2099 Hzbrightness

vocal movement · key A minor (confidence 0.77)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

video: not published yet — it goes here when it goes live on lives10

Se Paga (PT)

The theme as an anthem: no one tells the price; they only show the trophy. Female/male duet in rapid call and response.

constant intensity

style prompt sent to Suno

epic cinematic anthem, screaming female belting, raspy desperate female vocals, deep cavernous male baritone, viking folk duet, fast vocal call and response, gang vocal choir, orchestral war drums, soaring strings, raw aggressive vocals, 118bpm, huge chorus, brazilian portuguese

V5 model · style mode · female voice · styleWeight 0.88

clip history · clipe-v2 library · dark · medium cuts · blending and window

Two alternating layers: the arena where the duo sings, and the story of someone who wakes before sunrise, carries what no one wanted, and gets knocked back by closed doors. The double exposure is where the two meet without a cut. It ends with the arena lights slowly going out.

prompts used in the images and animation

finish applied to EVERY image in the project

raw documentary photograph, live-action cinema still, anamorphic 35mm, natural skin texture and pores, film grain, practical light, shallow depth of field, muted teal and amber grade

negative, also across all of them

cartoon, 3d render, cgi, pixar, anime, illustration, painting, plastic skin, doll face, oversaturated, text, watermark, extra fingers, deformed

animation prompt template (Agnes)

Smooth cinematic transition between the keyframes: <movimento da cena>. Natural motion, consistent characters and style, cinematic camera.

The 22 scene descriptions and the 22 camera movements in this library are at v2-cenas.json e v2-movs.json. The complete step-by-step guide is at docs/clipes.

Track 14:08

16.3 dBdynamics

5sustained

26 centsvibrato

2517 Hzbrightness

vocal movement · key D# minor (confidence 0.57)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

Track 23:57

16.4 dBdynamics

2sustained

—vibrato

2567 Hzbrightness

vocal movement · key A# major (confidence 0.49)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

video: not published yet — it goes here when it goes live on lives10

It's Paid (EN)

English version, made in cover mode from the PT track — inherits phrasing designed for Portuguese.

constant intensity

style prompt sent to Suno

epic cinematic anthem, screaming female belting, raspy desperate female vocals, deep cavernous male baritone, viking folk duet, fast vocal call and response, gang vocal choir, orchestral war drums, soaring strings, raw aggressive vocals, 118bpm, huge chorus, english lyrics

V5 model · cover mode · female voice · styleWeight 0.88

clip history · clipe-v2 library · dark · fast cuts · blending only

The same images as the PT version, but shuffled and with faster cuts so it doesn't become a clone. No window: here the overlay is always a double exposure.

prompts used in the images and animation

finish applied to EVERY image in the project

raw documentary photograph, live-action cinema still, anamorphic 35mm, natural skin texture and pores, film grain, practical light, shallow depth of field, muted teal and amber grade

negative, also across all of them

cartoon, 3d render, cgi, pixar, anime, illustration, painting, plastic skin, doll face, oversaturated, text, watermark, extra fingers, deformed

animation prompt template (Agnes)

Smooth cinematic transition between the keyframes: <movimento da cena>. Natural motion, consistent characters and style, cinematic camera.

The 22 scene descriptions and the 22 camera movements in this library are at v2-cenas.json e v2-movs.json. The complete step-by-step guide is at docs/clipes.

Track 14:43

14.8 dBdynamics

5sustained

82 centsvibrato

2291 Hzbrightness

vocal movement · key D# minor (confidence 0.49)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

Track 24:47

16.5 dBdynamics

7sustained

67 centsvibrato

2262 Hzbrightness

vocal movement · key D# minor (confidence 0.51)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

video: not published yet — it goes here when it goes live on lives10

Don't Give Up

Perseverance: the one who stands is not the one who never fell, but the one who came back the next day. First in the corrected batch.

corrected contrast

style prompt sent to Suno

raw emotional male and female duet, gravelly baritone spoken responses, driven chest mix, no falsetto, wide slow vibrato, laid back phrasing, extreme dynamic contrast, quiet intimate verse, huge belted chorus, orchestral build, live raw take, brazilian portuguese, 76bpm

negatives

muffled, lo-fi, constant screaming, flat dynamics, autotune, falsetto

V5 model · style mode · female voice · styleWeight 0.7

clip history · clipe-v4 library · intimate · medium cuts · natural light

Just one character, from her bedroom at five in the morning to a small club stage. No arena, no pyrotechnics: handheld camera, window light, empty rehearsal room. The progression is chronological and ends with her heading out into the early morning.

prompts used in the images and animation

finish applied to EVERY image in the project

raw documentary photograph, live-action cinema still, anamorphic 35mm, natural skin texture and pores, film grain, practical light, shallow depth of field, muted teal and amber grade

negative, also across all of them

cartoon, 3d render, cgi, pixar, anime, illustration, painting, plastic skin, doll face, oversaturated, text, watermark, extra fingers, deformed

animation prompt template (Agnes)

Smooth cinematic transition between the keyframes: <movimento da cena>. Natural motion, consistent characters and style, cinematic camera.

The 22 scene descriptions and the 22 camera movements in this library are at v4-cenas.json e v4-movs.json. The complete step-by-step guide is at docs/clipes.

Track 12:01

18.1 dBdynamics

13sustained

70 centsvibrato

1990 Hzbrightness

vocal movement · key E minor (confidence 0.73)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

Track 22:33

22.0 dBdynamics

31sustained

59 centsvibrato

2031 Hzbrightness

vocal movement · key E major (confidence 0.69)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

video: not published yet — it goes here when it goes live on lives10

The Center

Human talent in the face of superintelligence. The machine amplifies what already exists in a person; whoever arrives empty amplifies the emptiness.

corrected contrast

style prompt sent to Suno

epic cinematic anthem, powerful female belt, deep male spoken word, gang vocal choir, no falsetto, wide slow vibrato, quiet piano intro, massive orchestral chorus, war drums, extreme dynamic contrast, brazilian portuguese, 84bpm

negatives

muffled, lo-fi, constant screaming, flat dynamics, autotune, falsetto

V5 model · style mode · female voice · styleWeight 0.7

clip history · clipe-v5 library · graphic · slow cuts · centered text

Blue and amber, symmetrical and nearly abstract. The machine appears as a server room and network of cool light; the human appears as hands, faces, and the family table in warm light. Text is the main device because the song is a manifesto. It closes with dawn filling the entire screen.

prompts used in the images and animation

finish applied to EVERY image in the project

raw documentary photograph, live-action cinema still, anamorphic 35mm, natural skin texture and pores, film grain, practical light, shallow depth of field, muted teal and amber grade

negative, also across all of them

cartoon, 3d render, cgi, pixar, anime, illustration, painting, plastic skin, doll face, oversaturated, text, watermark, extra fingers, deformed

animation prompt template (Agnes)

Smooth cinematic transition between the keyframes: <movimento da cena>. Natural motion, consistent characters and style, cinematic camera.

The 22 scene descriptions and the 22 camera movements in this library are at v5-cenas.json e v5-movs.json. The complete step-by-step guide is at docs/clipes.

Track 12:52

24.7 dBdynamics

28sustained

31 centsvibrato

2083 Hzbrightness

vocal movement · key A minor (confidence 0.89)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

Track 22:45

27.9 dBdynamics

28sustained

30 centsvibrato

2092 Hzbrightness

vocal movement · key D major (confidence 0.68)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

video: not published yet — it goes here when it goes live on lives10

The Wheel

No one grows alone: recognize those who came before and help keep the wheel turning. Gospel rock with organ and clapping.

corrected contrast

style prompt sent to Suno

soulful gospel rock, raspy male lead, female harmony, full choir, hammond organ, handclaps, no falsetto, wide slow vibrato, quiet intro, explosive final chorus, extreme dynamic contrast, brazilian portuguese, 92bpm

negatives

muffled, lo-fi, constant screaming, flat dynamics, autotune, falsetto

V5 model · style mode · male voice · styleWeight 0.7

clip history · clipe-v3 library · modern · fast cuts · burst and split

Joyful and provocative: a band on a rooftop at sunset, a choir clapping in a brightly lit room, dancers in a painted alley, crowd surfing, and confetti. It's the only one with a burst (images at 0.28s) and split screen to match the tempo.

prompts used in the images and animation

finish applied to EVERY image in the project

raw documentary photograph, live-action cinema still, anamorphic 35mm, natural skin texture and pores, film grain, practical light, shallow depth of field, muted teal and amber grade

negative, also across all of them

cartoon, 3d render, cgi, pixar, anime, illustration, painting, plastic skin, doll face, oversaturated, text, watermark, extra fingers, deformed

animation prompt template (Agnes)

Smooth cinematic transition between the keyframes: <movimento da cena>. Natural motion, consistent characters and style, cinematic camera.

The 22 scene descriptions and the 22 camera movements in this library are at v3-cenas.json e v3-movs.json. The complete step-by-step guide is at docs/clipes.

Track 12:13

29.6 dBdynamics

12sustained

68 centsvibrato

2397 Hzbrightness

vocal movement · key C major (confidence 0.82)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

Track 22:26

33.7 dBdynamics

6sustained

—vibrato

2797 Hzbrightness

vocal movement · key C major (confidence 0.75)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

video: not published yet — it goes here when it goes live on lives10

Luz de Volta

Emotional rock in the style of the tears-v3a reference: quick voice changes with no breathing room, vocals in front of the band, and few doubles.

new

style prompt sent to Suno

emotional rock anthem, male and female lead trading lines with zero gap, tight interlocking call and response, instant overlapping vocal handoff, vocals mixed loud and upfront above the band, dry close vocal, minimal double tracking, single tracked lead, straight tone with almost no vibrato, high belt range, raw honest delivery, quiet intimate intro, extreme dynamic contrast, driving guitars, live drums, brazilian portuguese, 82bpm

negatives

heavy vocal doubling, stacked harmonies, choir pads, autotune, buried vocals, instrumental heavy mix, wide vibrato, flat dynamics, gap between vocal lines, constant screaming

V5 model · style mode · female voice · styleWeight 0.72

the clip

No video yet: the song was rejected and will be remade before becoming a video.

Track 12:53

15.5 dBdynamics

3sustained

—vibrato

2467 Hzbrightness

vocal movement · key E minor (confidence 0.72)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

Track 22:04

9.8 dBdynamics

5sustained

—vibrato

2192 Hzbrightness

vocal movement · key E major (confidence 0.91)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

video: not published yet — it goes here when it goes live on lives10

Collection — our previous songs

Tracks generated before this workflow, without the saved spec. They are included without the style prompt, but with audio, measurements, key, and the same charts—which makes comparison possible.

Tears Become Light

Emotional rock. The reference that became the target: breathes, builds without staying in a scream, straight and tense vocals.

collection

tears-v3a2:28

24.1 dBdynamics

16sustained

17 centsvibrato

2203 Hzbrightness

vocal movement · key A# minor (confidence 0.69)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

tears-v3b3:27

20.9 dBdynamics

11sustained

34 centsvibrato

2447 Hzbrightness

vocal movement · key A# major (confidence 0.85)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

We Were Here

The most 'held back' of the entire collection: 44 sustained notes, the most sustained notes among those generated.

collection

wewerehere-v2a2:31

24.4 dBdynamics

44sustained

37 centsvibrato

2514 Hzbrightness

vocal movement · key C major (confidence 0.93)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

wewerehere-v2b2:06

25.1 dBdynamics

32sustained

28 centsvibrato

2426 Hzbrightness

vocal movement · key A# major (confidence 0.86)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

Fire

Lower and spread across the registers, with little sustain.

collection

fire-v1a4:12

22.1 dBdynamics

16sustained

198 centsvibrato

2386 Hzbrightness

vocal movement · key G minor (confidence 0.78)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

fire-v1b3:28

23.8 dBdynamics

4sustained

—vibrato

2358 Hzbrightness

vocal movement · key G minor (confidence 0.83)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

Horizons

The one that spends the most time screaming: one of the tracks spends over 89% of its time in belt.

collection

suno-horizons-v1-chops2:19

17.3 dBdynamics

11sustained

66 centsvibrato

1278 Hzbrightness

vocal movement · key C major (confidence 0.89)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

suno-horizons-v2-chops3:06

23.1 dBdynamics

7sustained

41 centsvibrato

2228 Hzbrightness

vocal movement · key C major (confidence 0.91)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

Digital Sky

Sits in the upper register, medium to low dynamics.

collection

suno-digital-sky-v13:21

17.1 dBdynamics

5sustained

47 centsvibrato

1940 Hzbrightness

vocal movement · key C major (confidence 0.82)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

suno-digital-sky-v23:28

13.8 dBdynamics

13sustained

31 centsvibrato

2517 Hzbrightness

vocal movement · key C major (confidence 0.87)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

Builders

Low dynamics and almost no sustain—one of the tracks has ZERO sustained notes.

collection

suno-builders-v1-vocal3:16

13.7 dBdynamics

2sustained

43 centsvibrato

2234 Hzbrightness

vocal movement · key F major (confidence 0.88)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

suno-builders-v2-vocal2:49

15.1 dBdynamics

0sustained

—vibrato

2377 Hzbrightness

vocal movement · key F major (confidence 0.87)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

Isolated cover

The smallest dynamic range in the entire collection, and one of the tracks also has no sustained notes.

collection

suno-cover-isolado-v13:04

14.8 dBdynamics

0sustained

—vibrato

2501 Hzbrightness

vocal movement · key G major (confidence 0.85)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

suno-cover-isolado-v23:35

10.8 dBdynamics

16sustained

32 centsvibrato

2127 Hzbrightness

vocal movement · key C major (confidence 0.89)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

Cover reel

High dynamics in one track, almost no sustain in the other.

collection

suno-cover-reel-v12:14

28.2 dBdynamics

2sustained

—vibrato

2183 Hzbrightness

vocal movement · key A# major (confidence 0.84)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

suno-cover-reel-v22:52

15.1 dBdynamics

12sustained

90 centsvibrato

2380 Hzbrightness

vocal movement · key F major (confidence 0.81)

spectrogram (low notes below, high notes above; color = volume)

waveform (volume only)

piano roll (each block = one sung note)

dialogue between the voices (blue = register change)

notes used (amber = tonic)

where the voice spends its time

The references, measurements

Your computer's reference library, evaluated by the same standard. There is intentionally no player here: some of these recordings belong to third parties, and we don't have the rights to publish them. What matters are the numbers—they show why one sounds desperate and another sounds amateurish.

reference	key	median F0	dynamics	sustained	longer	vibrato	% belt
live human (AGT) REAL HUMAN VOICE: 182 sustained notes and 41 dB — no generated voice comes close	D minor	294 Hz	41.1 dB	182	3.1s	57 cents	19%
human, noisy recording (AGT) same origin, recording contaminated by band and audience	A# minor	286 Hz	14.9 dB	32	0.9s	54 cents	29%
TikTok audio downloaded externally, used only as a style reference	The smallest	227 Hz	59.1 dB	28	0.7s	83 cents	2%
viking-gods (ref) a clip downloaded from a link; generation never actually ran	C minor	310 Hz	9.4 dB	2	0.7s	—	33%

median F0 shows where the voice sits. Notice that almost all of Suno’s output clusters between 469 and 595 Hz, a high, generic range, while the two human voices are near 290 Hz—almost an octave lower. % belt shows whether the music lives at full volume: above 75%, there’s no midrange left, and without a midrange, there’s nowhere to build from. median F0 does not determine the gender of the singer — men in a high belt and women in the mid range overlap in this band.

The chart that sums it all up

Each dot represents a track. The farther to the right, the more the music breathes (dynamics). The farther toward up, the more the voice sustains the notes. The upper-right corner is where the music that moves you lives. Hover to see the name and numbers.

our songs generated references real human voice

The human voice is alone up there. 182 sustained notes and 41 dB. The best thing Suno has produced in its entire catalog has 44 sustained notes. It's not a matter of tone: it's that the human voice holds phrases and varies volume in a way the model rarely does. That's the real ceiling we're dealing with.

How to classify a voice

The minimum vocabulary for describing any voice, the number that measures each thing on this page, and the word that asks Suno for it. That's what turns "I liked that one" into an instruction you can repeat.

concept	what it is	where to see it here	how to ask
Tessitura	the range of notes where the voice is comfortable	median F0 and the “where the voice spends its time” chart	`low register`, `high belt range`
Log	the mechanism: chest (thick, strong), head (light), mix (in between), falsetto (breathy)	% in belt on the bar chart	`driven chest mix`, `no falsetto`
Belt	carry the chest weight into the high register—it’s the sung “shout”	440–700 Hz range in the bars	`emotional belting`
Vibrato	note oscillation: rate (Hz) and depth (cents)	“vibrato” metric on each track	`wide slow vibrato` or `straight tone`
Sustain	how long the voice holds a note	“sustained” metric and the straight line on the movement chart	`long sustained notes`
Attack	how the note starts: directly or sliding up from below (scoop)	jumps at the start of phrases in the motion graph	`scooped attacks`, `direct attack`
Texture	clean, raspy (grit), breathy, nasal	brightness in Hz — above 2400 it sounds harsh	`gritty rasp`, `clean tone`, `breathy`
Dynamics	the distance between the lowest and highest notes	“dynamic” metric and the gray movement line	`quiet intro` + `explosive chorus`
Phrasing	the timing: ahead, behind (layback), rushed	block size and spacing in the dialogue	`laid back phrasing`
Interaction	how the voices take turns and how much silence they leave	“dialogue” chart: go to the exchange and the blue blocks	`zero gap`, `overlapping vocal handoff`
Key	the key: the central note and whether it’s major (open) or minor (tense)	“key” column and the note histogram	`in D minor`

The recipe that works

After comparing all 15, this is the order that produced the best results. The common mistake is to ask only for the genre and expect the voice to come along with it.

1. Genre first, and be specific. cinematic pop rock, power ballad carries more information than rock. Genre determines the instrumentation, tempo, and even the type of voice the model chooses.

2. Name the calm section. quiet piano intro was the most effective instruction of all the ones we tested—more than asking for “dynamic contrast,” because it names the instrument and the moment. Without a quiet section, there is no explosion.

3. Describe the vocal mechanism, not the emotion. driven chest mix, no falsetto, wide slow vibrato works; “emotional voice” doesn’t. The model understands vocal technique better than adjectives.

4. Say how the voices come together. zero gap, instant overlapping vocal handoff. Without this, the model leaves almost a second of silence between the call and the response, and the duel effect is lost.

5. Negative tags for what you don't want. constant screaming, flat dynamics, childish voice, thin vocals. That's what prevented the flattening of the first ones from happening again.

6. styleWeight 0.85 or higher when character matters. The only song we made with 0.72 was the one that came out without personality.

7. Explicit language. english lyrics or brazilian portuguese — the model doesn’t infer it from the rest of the prompt.

What we learned by comparing

Everything below came from comparing the prompts and measurements on this page.

Constant intensity falls flat. The first three songs called for screaming, desperate e raw aggressive at the same time, with no marking of calm sections. Result: 16 dB of dynamics and two sustained notes throughout the entire song. It sounds big and doesn’t move you because it never leaves the ceiling.

Naming the calm passage is the most efficient instruction. More direct than asking for "dynamic contrast": quiet piano intro yielded the best phrase sustain in the batch, and quiet intro along with explosive final chorus yielded the greatest contrast because it marks the beginning and end.

Negatives keep things from going overboard. Add constant screaming e flat dynamics in negativeTags was what prevented repeating the mistake from the first attempts.

Cover mode inherits the phrasing. The English version was made as a cover of the Portuguese track, so it was stuck with a rhythmic pattern designed for Portuguese.