Suno Prompts With Real Examples

Put your headphones on — we're about to break down Suno prompts with real examples of the tracks they produced. The article has 9 real examples with prompts and lyrics markup: different genres and techniques, plus an attempt to recreate the style of favorite artists and composers — using a soundtrack in the spirit of Hans Zimmer as the example. Be sure to listen to the experimental track — it shows what you can do with song lyrics.

First, a complex example that shows what Suno is capable of. A five-part rock opera in a single track: a piano ballad, a cappella, operatic call-and-response between the lead singer and a choir, hard rock and a quiet ending. That doesn't happen by accident — only through the prompt and the lyrics markup. Both takes from one generation:

The Last LighthouseTake 1 · rock opera · 4:33The Last LighthouseTake 2 · rock opera · 4:48

The style — the song's whole dramaturgy, in order:

Multi-part symphonic rock opera suite, theatrical and dramatic, about five minutes. Opens as a slow rubato piano ballad with an intimate male tenor, soft strings entering in the first verse. Then a true a cappella section: all instruments stop completely, only stacked multi-part vocal harmonies, dry and close. Then an operatic middle section: theatrical call-and-response between the lead vocal and a choir, timpani hits, sudden dynamic shifts and tempo changes. A short soaring melodic guitar solo launches a fast hard rock section with driving drums, heavy distorted guitars and powerful belted vocals, gang vocals on the repeats. Ends with a sudden drop back to solo piano and a soft voice, a final piano chord and a soft gong. Big dynamic contrast, every section clearly different, the same lead singer throughout, rich stereo layering, 1970s-inspired production with modern clarity. English lyrics

And this is the skeleton of the lyrics markup — each part is labeled with what should happen in it:

[Intro - solo piano, soft male tenor, slow and free]
[Verse 1 - piano ballad, intimate, soft strings enter]
[A Cappella - voices only, all instruments stop]
[Operatic Section - choir call-and-response, timpani, sudden dynamic shifts]
[Guitar Solo - short soaring melodic lead, full band enters]
[Hard Rock - fast driving drums, heavy guitars, powerful belted vocals]
[Hard Rock - even faster, gang vocals on the repeats]
[Final Section - sudden drop to solo piano, slow, soft male voice]
[Outro - final piano chord, soft gong]
[End]

All five parts are there: the ballad, the call-and-response with the choir, hard rock that's noticeably faster and heavier, and a return to piano at the end. One honest caveat: the a cappella didn't come out fully clean — quiet music remains under the voices. Why, and what to do about it, is in the metatags section. Next, step by step: what prompts like this are made of.

How a Suno prompt works

Suno has two modes, and a prompt means something different in each.

In Advanced mode in GPTunneL, the style field holds up to 1,000 characters and the lyrics field up to 5,000. Under the style field there are quick chips for genres and voice gender. Don't max out the lyrics limit, though: according to HookGenius measurements, past roughly 3,000 characters Suno starts cramming the song. There are two versions — v6 and v6 mini; a generation costs $0.24 and gives you two tracks at once.

The formula: the layers a prompt is built from

A good style prompt is built from six layers, and the order isn't random: according to HookGenius observations, Suno leans harder on whatever comes first in the description. So genre goes first and mix details go last.

A weak prompt looks like this: "a happy summer song". A strong one looks like this: "summery indie pop, around 110 BPM, acoustic guitar and handclaps in the verses, live drums come in on the chorus, bright female vocal close to the mic, warm vintage sound". The second gives the model something to hold on to in every layer, and the result stops being a lottery.

Lyrics markup: metatags

Metatags in square brackets tell Suno the structure: where the verse is, where the chorus is, where the break goes. The tags in the table appear in the tracks in this article — some we placed, some Suno placed itself.

The full dictionary with vocal and instrumental tags is in the Suno guide.

Behavior right in the tag

A tag on its own only says where a part is. What happens in it is better added after a hyphen: [Final Chorus - bigger, layered harmonies]. The difference was clearly audible in our tests. In a ballad generated for this article, the final chorus tagged exactly like that sounds bigger than the regular ones. In "The Dawn That Stays" the final chorus differs only by new lines — and came out just slightly stronger. The takeaway: Suno won't invent an abstract lift on its own — you have to name it.

Parentheses — a different voice

Suno sings text in parentheses in a separate voice, like a backing vocal or an echo. In "The Dawn That Stays" — that track is a little further down, in the hierarchy section — the line "(We are still here)" in the intro and outro sounds exactly like that. The same song shows lyrics and style working together: "Hey!" at the start of the chorus plus "forceful shouts" in the style — and Suno shouts it.

Breakdown: the rock opera from the intro

Now you can see how the track the article opened with is built. The style isn't a list of features but a script: the song's parts are described in order, each with its own instruments, tempo and voice.

The lyrics markup repeats the script part by part — tags say where each part is, and the behavior in the tag says what happens in it:

The call-and-response works through parentheses: the lead sings a line and the choir answers with the same line in parentheses — "Who will guide them in? / (Who will guide them in?)". The full lyrics — to recreate the track or borrow its structure:

[Intro - solo piano, soft male tenor, slow and free]
Seven nights the storm has stayed,
Seven lamps I've lit in vain.

[Verse 1 - piano ballad, intimate, soft strings enter]
Keeper of a broken light,
Counting ships that pass at night.
Every one I couldn't save
Sings to me beneath the waves.

[A Cappella - voices only, all instruments stop]
Hold the light, hold the light
Through the dark and through the night

[Operatic Section - choir call-and-response, timpani, sudden dynamic shifts]
Who will guide them in?
(Who will guide them in?)
Not the sea, not the wind!
(Not the sea, not the wind!)
Ring the bell, let the tower sing!
(Let the tower sing!)
I will not let the darkness in!
(Never let the darkness in!)

[Guitar Solo - short soaring melodic lead, full band enters]

[Hard Rock - fast driving drums, heavy guitars, powerful belted vocals]
Break the glass and burn the oil,
Every beam against the storm!
Fire in my hands tonight,
I'm the last lighthouse, I'm the light!

[Hard Rock - even faster, gang vocals on the repeats]
Burn it high! (Burn it high!)
Burn it bright! (Burn it bright!)
Till the morning finds the shore,
I will hold the light!

[Final Section - sudden drop to solo piano, slow, soft male voice]
The sea is calm, the ships are home,
The lamp still burns, I'm not alone.

[Outro - final piano chord, soft gong]
(Hold the light)

[End]

Experimenting: text as rhythm, voices and stereo

Who said Suno can only generate worn-out motifs? Listen to this experimental track, where the main material is a single word. "Loop" breaks into syllables and flows into "loose", "lose" and "love":

Loop Me Loose

Experimental art pop · Advanced mode · 2:53

0:002:53

Style promptsyllables as rhythmone word morphssplit stereo voicestwo-voice call and response

Experimental art pop built from a single vocal motif, around 118 BPM, steady pulse. Starts with soft felt piano and one dry close-miked human voice. Step by step the syllables of the lyrics turn into rhythm: chopped consonants, stuttered repeats, pitch-shifted vowels and looped phonemes become the percussion and synth lines. The opening piano phrase returns as the bass line and chords. Two voices trade lines in the second verse and split left and right in the second chorus. Bridge breaks words down to single sounds, followed by an almost silent pause. Final chorus brings the clean human voice and the chopped vocal textures together into one big, clear hook. Emotional, catchy and precise, controlled and musical, no random noise or glitch chaos. English lyrics

The fragments of the markup the track rests on:

[Pre-Chorus 2 - call and response]
[Voice A]
REPEAT ME
[Voice B]
remember me

[Chorus 2 - split stereo voices]
[Left Voice]
LOOP LOOP LOOP
[Right Voice]
LOSE LOSE LOSE
[Center Voice]
LOOP ME LOOSE

[Bridge - phonetic breakdown]
LOOP
LOOSE
LOSE
LOVE
L-L-L-L

The key idea of this example: the style sets the tempo, and the lyrics set the vocal rhythm. The style says 118 BPM and a steady pulse. How densely the voice sounds inside that grid depends on how the lyrics are written.

• Syllables with hyphens — L-L-LOVE, re-re-remember — are sung in chops, each piece on its own beat, and the voice turns into a percussion instrument. It worked.
• One word per line — LOOP / LOOSE / LOSE / LOVE — gives a sparse delivery and an audible transformation of the word. It worked.
• The [Left Voice] and [Right Voice] tags split the voices into the left and right channels — listen on headphones. It worked, and it's a rare confirmation of stereo placement through markup.
• The [Voice A] and [Voice B] tags produced different voices, but very similar ones. For more contrast, describe the timbre right in the tag: [Voice A - low male], [Voice B - high airy female]. We haven't tested this.
• Caps versus lowercase were meant as loud chopped moments against quiet ones. There's an audible difference, but a small one.

How metatags work in general is covered in the prompt-structure section above; more markup techniques on real tracks are in "Metatags in practice" below.

Hierarchy beats detail

There are two extremes. The first is "tag soup": genres and adjectives separated by commas with no priorities. The second is a wall of instructions where every instrument, bar and effect is spelled out with equal weight. What works is the middle: a few decisive instructions, each of which changes a specific musical decision.

Here's a good example — metalcore in Advanced mode, with all three fields filled in:

The Dawn That Stays

Metalcore · Advanced mode · 4:37

0:004:37

Style promptcinematic metalcorehalf-time chorus hitssub-drop impactsmelodic chorus lifts

Title:

The Dawn That Stays

Style:

Cinematic metalcore with melodic hardcore, driving mid-fast pulse and half-time chorus hits, sub-drop impacts punctuating transitions, synth string pads beneath downtuned guitars and pounding drums, aggressive male vocals with forceful shouts and melodic chorus lifts

[Intro]
(We are still here)

[Verse 1]
Neon rain on an empty street
The whole world hums beneath our feet
We chased a voice through the underground
Lost its signal, kept its sound

The towers blink like distant stars
We count the light between the dark
Every road they drew for us
Ends where we begin

[Pre-Chorus]
Hear it rising through the wires
One small spark becomes a fire

[Chorus]
Hey! We are still here
Calling out above the noise
Let the skyline disappear
We will follow our own voice
When the last light burns away
We'll be the dawn that stays

[Verse 2]
The station clocks have all gone still
But we can hear the engines build
Names we buried in the rain
Come back singing through the train

We take the stairs into the storm
With shaking hands, with hearts still warm
The city cannot tell us now
Who we're meant to be

[Pre-Chorus]
Hear it rising through the wires
One small spark becomes a fire

[Chorus]
Hey! We are still here
Calling out above the noise
Let the skyline disappear
We will follow our own voice
When the last light burns away
We'll be the dawn that stays

[Bridge]
If the night swallows every sign
Take my hand and I'll take mine
From every voice along the way
We make a road into the day

[Final Chorus]
Hey! We are still here
Louder than we've ever been
Let the skyline disappear
Watch the morning breaking in
When the last light burns away
We'll be the dawn that stays

[Outro]
(We are still here)
The rain lets go
The streets turn gold
We are still here

The style is about forty words, and almost every one changes a specific musical decision:

Two techniques from here are worth taking with you. Tie events to the structure — "in the chorus", "at the transitions" — and the model gets a script, not a pile of sounds. And use relationships instead of lists: "pads beneath guitars" says more than "pads, guitars".

And two numbers from this example. The style takes only 267 of 1,000 characters — there's no need to use the whole limit. Lyrics of 1,488 characters produced a song of 4 minutes 37 seconds.

Your favorite band's or composer's sound: a soundtrack in the spirit of Hans Zimmer

An artist's name in a prompt doesn't work as a recipe: Suno ignores or substitutes it. But the sound people love a band for can be broken down into parts, and each of them can be named in words:

• Era — "late 90s", "early 2000s".
• Rhythm and tempo — straight or swung, half-time, roughly how many BPM.
• Instruments and their roles — what leads, what supports, what adds texture.
• Vocals — register, timbre, delivery: rap, clean singing, strain.
• Production — raw or glossy, dry or spacious, dense or airy.
• Structure — how fast the hook arrives, what makes the parts contrast.

Here's how that sounds on a real track — a soundtrack in the spirit of mid-2010s space sci-fi, without a single composer or film name:

The Long Orbit

Film score · Advanced mode · 2:29

0:002:29

Style promptchurch pipe organone repeating motifslow layered buildno trailer drums

Cinematic space-epic film score in the minimalist style of mid-2010s Hollywood science fiction, instrumental, around 70 BPM, slow and patient. Built on one short repeating motif played on a huge church pipe organ, with soft felt piano doubling it an octave higher. A steady ticking pulse, like a clock, runs underneath the whole track. Starts almost empty: a single sustained organ chord and quiet piano. Layers are added one by one: low sustained strings, then pulsing organ arpeggios, then full organ with swelling strings and brass. Builds very slowly to an overwhelming, emotional climax where the organ fills the whole space, then drops back to piano and a fading organ chord. Vast cathedral-like reverb, deep sub-bass, wide stereo, dynamic range from near silence to massive. No drums, no trailer hits, no braams, no choir, no vocals

The lyrics field contains only [Instrumental]. Here's what each part is responsible for:

What worked: the organ leads, the track builds slowly, layer by layer, to a climax, and there are no drums or trailer hits. What worked weakly: the ticking pulse isn't obvious by ear. The detail that holds the whole style together deserves more weight — name what plays it and put it closer to the start of the prompt: "a clearly audible ticking pulse on muted piano from the very first second". One more observation: of the two takes from the generation, one came out noticeably closer to the idea and the other sounded generic. For complex styles like this that's normal — there's a reason you get two takes.

The same approach works for any band. The sound of early-2000s nu-metal without a single name:

Late-90s nu-metal with electronic textures, 95 BPM, heavy half-time groove. Down-tuned palm-muted guitars and a tight kick-snare pocket drive the verses; turntable scratches and a synth pad add texture; the chorus opens into wide distorted guitars. Male vocal switches between rhythmic rap delivery in the verses and a clean melodic chorus with strained emotional edge. Bridge drops to filtered vocal and beat, then the final chorus returns with layered harmonies. Dense live mix, audible bass, no glossy over-polish. English lyrics

That way the result is closer to what you had in mind — and stays yours, not a copy of someone else's song.

Metatags in practice

How metatags behave on real tracks — from a greeting in Simple mode to an instrumental without a single word, and the places where Suno reaches its limits.

A name in the chorus and tags from Suno

For a birthday song in Simple mode, one sentence was enough: in our test for this article, Suno wrote the lyrics itself, put the friend's name in the chorus and placed the tags. A prompt built the same way:

Warm acoustic pop birthday song for my friend Emma: female vocal, ukulele and light percussion, cheerful catchy chorus with her name, English lyrics

Your own lyrics in Simple mode

You can pass your own words even without Advanced mode — right inside the description, with the command "Sing exactly these lyrics":

Style: indie pop, female vocals, acoustic guitar, soft drums, 100 BPM. Sing exactly these lyrics:
[Verse]
Morning smells of coffee and the rain,
We overslept together once again.
[Chorus]
Don't rush, the city waits for us,
This day has only just begun.

In a test for this article, the lyrics Suno returned matched the submitted text line for line, and it took the song title from the chorus. Four lines make about a minute of music: a full song needs at least two verses, a chorus and a bridge.

Structure without lyrics: light music for an ad

In an instrumental, tags have nothing to hold on to — there are no words. Then the structure is set right in the style: when the main melody appears, where the lift is, how it ends. Music for a commercial or a summer café background:

Sunlit Café

Instrumental for an ad · Advanced mode · 1:58

0:001:58

Style promptimmediate hookwhistled melodygentle lift in the second halfclean warm ending

Sunlit indie pop, breezy bossa nova groove, 104 BPM, warm nylon-string guitar, playful muted guitar accents, mellow Rhodes piano, round melodic bass, softly brushed drums, hand-played shakers, light handclaps, natural human timing, catchy whistled melody, short repeating hook, relaxed forward motion, subtle syncopation, airy spacious mix, gentle tape warmth, cheerful and carefree, summer café atmosphere, understated feel-good energy, instrumental, immediate hook, small melodic variations, gentle lift in the second half, clean warm ending

The lyrics field contains only [Instrumental]. What worked: the whistled melody comes in right away, and the track ends cleanly — easy to cut for a video. For an ad, those are the two main requirements. What worked weakly: the lift in the second half isn't obvious by ear. Once again it's an abstract "lift" — if you need a clear one, name what changes: "in the second half, handclaps and a second whistled harmony join, drums switch from brushes to a fuller beat".

The prompt is long, and the mood repeats itself: "cheerful and carefree", "feel-good energy" and "summer café atmosphere" say almost the same thing. Any two of the three can go without loss.

Where the limits are: silence and a cappella

Suno doesn't deliver complete silence from the instruments. In the first version of the rock opera from the start of the article, the a cappella was tagged [Harmonies - a cappella, no instruments] and its lines were in parentheses — and light music played under the voices. The second version fixed three things at once:

[A Cappella - voices only, all instruments stop]
Hold the light, hold the light
Through the dark and through the night

The tag is named directly, and the lines have no parentheses — since parentheses mean backing vocals, and backing vocals imply accompaniment. The style gained "all instruments stop completely… dry and close". The section became more expressive, but the background under the voices stayed. The same happened with the [Breakdown - almost silent] tag in "Loop Me Loose" above: instead of near silence, a quiet piano plays there. The takeaway: these edits make such moments more vivid, but a prompt doesn't guarantee complete silence.

Vocals and pronunciation

Suno sings English lyrics confidently — the tracks above prove it. The weak spots are words that can be read more than one way, plus numbers and abbreviations. A few habits that lower the risk:

• Say the language explicitly. "English lyrics" or "lyrics in Spanish" — otherwise Suno may pick the language on its own.
• Spell out numbers and abbreviations. "Twenty-five", not "25"; "deejay", not "DJ".
• Swap out words with two pronunciations. "Read", "live" and "lead" sound different depending on meaning; a clear synonym or a rephrased line is the most predictable fix.
• Keep lines short and the meter even — that makes it easier for the model to fit the words to the melody.
• Write section tags in English. They're tested on every track in this article.

Other languages work too. For a ballad in another language generated for this article, the combination of "style in English, lyrics in the song's language" gave correct stress and no accent. Where stress is ambiguous, capitalizing the stressed vowel helped in our test. And Suno understands markup in other languages: for a lullaby described in another language, it placed the section tags in that language itself.

If a word still comes out wrong, don't rewrite the whole song: replace that one line and run the generation again.

Write the essentials explicitly — Suno fills in the rest: a hip-hop example

Before generating, Suno expands your description into a detailed style and adds what wasn't there. The prompt for this hip-hop track is a single sentence:

One Floor at a Time

Hip-hop · Simple mode · 3:23

0:003:23

Style promptboom-bapmale rap flowdusty drum breakjazzy piano sample

Confident boom-bap hip-hop track about chasing a dream in a big city: male rap vocals with a sharp rhythmic flow, dusty drum break, deep bass, jazzy piano sample, catchy chorus hook, English lyrics

And here's what Suno expanded it into before generating:

confident golden-era boom-bap hip-hop, sharp rhythmic male rap delivery, dusty chopped drum break, deep rounded bass, jazzy piano sample with warm tape texture, crisp kick and snare, subtle turntable cuts, catchy sung-rap chorus hook, focused street-level storytelling, punchy dry vocal mix, dynamic verse-to-hook arrangement

The turntable cuts, the warm tape texture, the crisp kick and snare and the dry vocal mix weren't in the prompt — the model added them on its own. The practical takeaway: spell out everything that truly matters for the song, and Suno will fill in the rest. Sometimes that works out, sometimes it doesn't — and then change one parameter in the next generation instead of rewriting the whole prompt.

Voice as an instrument

A voice is described the same way as an instrument: register, timbre, distance from the mic and, above all, how the delivery changes over the course of the song. In a rock test for this article, a one-line prompt with "a strained, emotional male vocal and a building chorus" was expanded by Suno into section-by-section dynamics: "tense and restrained verses, a chorus that gradually builds into a wide anthemic burst". It's better to set that kind of dynamic yourself rather than rely on the model:

Alternative rock about coming home after a long time away: raspy male baritone, restrained and close in the verses, powerful and open in the chorus without shouting, driving electric guitars, live drums, backing harmonies only on the final chorus, English lyrics

What else works with vocals:

• Voice gender — in words in the style, and with the voice-gender chips in Advanced mode.
• Register and timbre — not "emotional vocal" but "warm low alto", "raspy baritone", "airy soprano".
• Backing vocals where they belong — "backing harmonies only on the final chorus" instead of a general "with backing vocals".
• Stable tone on long tracks. v6 users complain that the voice sometimes changes toward the end of a song. The phrase "keep vocal tone and loudness stable across the whole track" is a community tip; we haven't tested it separately.

Tempo, groove and "no vocals"

For Suno, a BPM number is a hint, not a command. What works more reliably is the number together with a description of how the rhythm feels: "around 124 BPM, steady four-on-the-floor beat, continuous eighth notes in the bass". If the model ignores the number, it still has a second route to the right feel.

Atmosphere comes from textures, not adjectives. Lo-fi for studying:

Warm F Piano

Lo-fi · Simple mode · 2:43

0:002:43

Style prompt75 BPMvinyl cracklerainno vocals

Calm lo-fi hip-hop for studying, no vocals: warm piano, soft drums, rain and vinyl noise, 75 BPM, instrumental

The prompt is translated from the original description the track was generated from. "Rain and vinyl noise" added more atmosphere than one more "cozy" would have. And "no vocals" worked literally: the lyrics of both takes contain nothing but the [Instrumental] tag. In Advanced mode, the same result comes from [Instrumental] in the lyrics field plus the word instrumental in the style.

If you'd rather not set BPM as a number, classical tempo markings come in handy — the model understands them as a guideline too:

Tempo build: a garage rock example

Real acceleration from one number to another is something Suno does unreliably. But the feeling of acceleration comes together from several techniques at once. This garage rock track is designed to speed up from start to finish:

White Noise

Garage rock · tempo build · 2:26

0:002:26

Style prompt100 → 140 BPMeighths to sixteenthsdouble-time final chorusno breakdown

Experimental folk garage rock, lo-fi garage, raw live-band take that keeps accelerating from start to finish: starts as a midtempo push around 100 BPM and speeds up steadily to about 140 BPM by the final chorus, never slowing down. Humanized drum kit with imperfect timing; hi-hats move from eighths to sixteenths, drums add fills before every chorus, final chorus in double-time. Hand-played acoustic guitar with raw amp breakup, off-grid tambourine, room mic bleed, tape saturation, spring reverb, clipped transients, mono punch, short collage edits that never stop the groove. Anxious momentum, caught in motion, ragged and adventurous. Vocals in every section, shouted gang vocals on the repeats. Very short intro, no instrumental breaks, no breakdown, no half-time, no fade-out, ends abruptly at peak energy. English lyrics

In the lyrics, the tags reinforce the same thing toward the finale:

[Build - drums accelerate, shouted vocals, no instrumental]
Turn it up, turn it up
Turn it up, turn it up
Louder, louder
Faster, faster

[Final Chorus - double-time, fastest, full energy]
White noise! White noise!
In the car, in my bones
White noise! White noise!
Turn it up, take me home
White noise! White noise!
Turn it up!

[End]

What worked: the tempo builds over the song, the final chorus is noticeably faster and denser than the rest, there are no long wordless stretches, and the track cuts off at its peak. What didn't fully work: there are no slow parts, but in places the tempo sags slightly — a prompt doesn't guarantee a smooth acceleration curve.

The acceleration is built from five parts:

• Numbers: "from 100 to 140 BPM" — they set the direction.
• Rhythmic density: hi-hats move from eighths to sixteenths, with fills before every chorus. The model hears this better than numbers.
• Double-time in the finale — the drums play twice as densely, and it sounds like speeding up.
• Bans on slowing down: no breakdown, no half-time, no fade-out.
• Shorter lines toward the end — "Louder, louder / Faster, faster": the vocal gets denser.

Genre, era and live sound

A single era word pulls in a whole set of sounds. "Retro 80s" in the prompt for this synthwave track means analog synths, gated drums and the characteristic sound of the mix:

Midnight Avenue

Synthwave · Simple mode · 4:24

0:004:24

Style promptretro 1980sanalog synthssynth solo before the final chorus

Retro 80s synthwave song about driving through a neon city at night: female vocals, analog synths, punchy drums, dreamy catchy chorus, English lyrics

In the expanded style, Suno added "instrumental synth solo before the final chorus" — a solo before the final chorus, a typical move of the era. The model knows genre conventions and applies them on its own if you name the genre and the period precisely.

If you mix genres, give them roles instead of listing them with commas. Not "industrial, trip-hop, doom metal, dark", but "trip-hop sets the slow broken beat and sub-bass, doom metal brings the heavy guitars, industrial adds the texture". That way the model knows what to take from where.

For a live sound, users suggest adding "live performance", "live audience" or "unplugged session" — as if the song were recorded at a concert. That's a community tip; we didn't use it in the tracks for this article.

Verbs instead of adjectives

The strongest technique of all: don't list which sounds are in the song — say what each one does and when. Compare two descriptions of the same idea. The first is a list of features:

alternative pop, electropop, 118 BPM, energetic, hopeful, airy female lead vocals, ambient pad textures, bright melodic synthesizer leads, pulsing analog bass synth, English expressive female soprano, clear and intimate with light reverb female vocals, crisp high-end EQ, polished vocal production; tight four-on-the-floor groove, rising synth layers from verse to chorus, wide pads, punchy kick and snare, memorable topline, clean modern master

The second is the same genre and tempo, but it describes behavior:

Alternative electropop, 118 BPM, tight four-on-the-floor pulse. Analog bass synth pulses in steady eighths under the verses; pads stay narrow and low until the pre-chorus, where synth layers rise one by one; bright melodic lead answers the vocal only in the chorus; punchy kick and snare enter fully at the hook. Airy female soprano, close and intimate in the verses with light reverb, opening into sustained confident notes in the chorus without shouting. Bridge strips to voice and one pad, then a filtered build returns the full arrangement. Vocal-forward modern master, wide chorus image, crisp high end. English lyrics

The difference is in the words "when" and "what it does". The instruments get entry points, the vocal gets a change of delivery by section, and the song gains a bridge and a return. A list of features gives the model a palette; a description of behavior gives it a script. The first approach works too, but only the second lets you control how the song develops.

Checklist before you generate

1. Genre and era — one main genre, a second one through "with" or with a role.
2. Pulse — BPM plus a description of the groove: straight, swung, half-time.
3. Players — what plays and when it comes in, not just a list.
4. Vocals — gender, register, timbre and how the delivery changes by section.
5. Arc — what changes toward the chorus, in the bridge, in the finale. An abstract "lift" doesn't work — name specifics.
6. Mix — one or two phrases about the sound.
7. Language — "English lyrics" or the song's language; for other languages, check words with ambiguous stress.
8. Structure and ending — tags with behavior, [Outro] or [End], so the track doesn't cut off mid-phrase.

A style prompt can be up to a thousand characters, but the good prompts in this article fit into 250–900. If something's off after generating, change one item at a time — otherwise you won't know what actually helped.

Ready-made prompts by genre

The prompts tested by generation are in the sections above, next to the tracks. The ones below follow the same formula — swap in your own topic.

Hip-hop and R&B

Dark trap about a city at night: confident male rap, heavy 808 bass, fast hi-hats, minor-key synths, short catchy hook, English lyrics

Smooth R&B about a late-night phone call: velvety female vocal, electric piano, unhurried groove, muted bass, intimate warm mix, English lyrics

Laid-back jazz-hop for the evening: relaxed male rap, saxophone sample, soft drums with vinyl hiss, 85 BPM

Upbeat hip-hop for a sports video: energetic rap, heavy kick, scratches, shouted hook, 95 BPM

Electronic

Melodic deep house for a summer terrace: pulsing bass, warm pads, light female vocal in short phrases, 122 BPM

Driving techno track, no vocals: hard steady kick, hypnotic arpeggiated synth, industrial textures, 130 BPM

Airy chillout for meditation: long pads, ocean sounds, sparse chimes, no drums, very slow tempo

High-energy EDM with a big drop: building intro, bright lead synth, massive drop, male vocal in the verses, 128 BPM

Rock and alternative

90s grunge about the old neighborhood: raspy male vocal, heavy distorted guitar, simple beat, guitar solo before the last chorus

Pop-punk about the last summer of school: fast tempo, bright guitars, cocky male vocal, gang-vocal chorus, live drums

Road-trip indie rock: light overdrive, guitar riff up front, husky female vocal, quiet verses and a wide chorus

Post-rock, no vocals: long delay-soaked guitar swells, slow build, explosive climax, cinematic ending

Lo-fi and background

Lo-fi for morning coffee, no vocals: lazy guitar, brushed drums, cozy record crackle, 70 BPM

Background ambient for work, no vocals: soft synth pads, a barely-there pulse, no sharp sounds or tempo changes

Jazz café, no vocals: upright bass, piano, brushes on the snare, easy swing, live chamber sound

Pop and acoustic

Summery indie pop about a trip to the sea: carefree mood, acoustic guitar and handclaps, bright female vocal, warm vintage sound

2000s dance pop: bright synths, tight beat, confident female vocal, a chorus you want to sing along to, 118 BPM

Acoustic breakup ballad: male vocal, a single guitar, soft and close to the mic, strings in the second verse

80s disco-pop about first love: soft synths, disco bass, male-female duet, bright nostalgic chorus

Piano ballad: tender female vocal, piano and cello, slow tempo, climax in the final chorus, English lyrics

Film and orchestral

Epic trailer music, no vocals: powerful brass, big taiko drums, choir at the climax, build toward a final hit

Fantasy game soundtrack, no vocals: strings and flute, Celtic motifs, bright adventurous mood, orchestral sound

Tense thriller music, no vocals: low drone, dissonant strings, sparse hits, a sense of unease

Moving theme for a film's ending, no vocals: piano, string orchestra, slow build, hopeful finale

Prompts for a specific task

Here the genre is secondary: what matters more is who the song is for and where it will play.

Birthdays and celebrations

Upbeat pop song for colleague Mike's work anniversary: bouncy male vocal, horns and handclaps, chorus with his name, playful lyrics about work and vacations

Tender wedding song: male-female duet, acoustic guitar and strings, slow tempo, lyrics about the first meeting

New Year's song for a kids' party: sleigh bells, cheerful choir, simple lyrics about the tree and presents, fast happy tempo

Ads and jingles

Background music for a video with a track is above — "Sunlit Café". For a jingle with a slogan:

Short radio jingle for the Bean There coffee shop: upbeat funk, female and male backing vocals, slogan "Bean There — mornings start here", about 30 seconds

While this article was being prepared, a jingle prompt asking for "about 30 seconds" came out at 28 and 30 seconds. Suno keeps length as a guideline, but don't count on exact timing for an edit.

Intro for a tech podcast: short electronic theme, pulsing bass, bright synth, no vocals, about 15 seconds

Jingle for an online flower shop: light acoustic pop, female vocal, slogan "Flowers at your door in an hour", about 20 seconds

Lullabies and kids' songs

Gentle lullaby for a baby: soft female vocal, piano and music box, slow tempo, simple tender lyrics about stars and sleep

Fun kids' song about dinosaurs: playful vocal, ukulele and xylophone, simple repeating chorus, fast tempo

Morning exercise song for kids: upbeat children's choir, simple instructions in the lyrics, drums and handclaps, 110 BPM

Background music for video and social media

Energetic background for short vertical travel videos, no vocals: pop-electronic, bright synths, handclaps, a short build, 120 BPM

Cozy background for a cooking video, no vocals: acoustic guitar, ukulele, light percussion, carefree mood

Intro for a gaming YouTube channel, no vocals: chiptune with a modern beat, 8-bit melodies, punchy opening, about 10 seconds

Team and corporate events

Sales team anthem for a company party: upbeat pop-rock, male and female choir, a chant-style chorus with the team's name, playful lyrics about targets and deadlines

Heartfelt farewell song for a colleague: acoustic guitar and piano, warm male vocal, lyrics about shared projects and good wishes for the new job

If something goes wrong

Common mistakes

• A pile of adjectives. "Epic, powerful, emotional, beautiful" — the model can't tell what matters most. Describe what makes the song powerful, and when.
• A list of instruments without roles. "Piano, guitars, strings, drums" — better "strings come in only on the final chorus".
• Contradictions. "Calm aggressive metal" forces the model to choose, and it doesn't always choose what you had in mind. Split different deliveries by section.
• Artist names. They don't work — break the sound down into parts.
• Lyrics without markup. Without [Verse] and [Chorus], the model decides on its own where the chorus goes, and often gets it wrong.
• Abstract development. "A lift in the second half", "a bigger finale" — Suno barely hears instructions like that. Name what exactly changes.
• Believing in precise control. Exact chords, bars and length down to the second can't be set with a prompt: a description is a guideline, not a score.
• Five changes after one miss. Change everything at once and you won't know what helped. One change at a time.

How to write song lyrics with AI

Suno writes lyrics decently on its own, but the words tend to come out generic — the kind that fit any occasion. If the song needs names, jokes and specific stories, it's easier to write the lyrics in a chat model and paste them into Advanced mode. The Suno page in GPTunneL has a Claude AI button right under the fields that writes lyrics in seconds. Or use a prompt for any text model:

Write song lyrics in English. Topic: [what the song is about, who it's for, which details to mention].
Genre and mood: [for example, upbeat pop].
Structure: [Verse 1], [Pre-Chorus], [Chorus], [Verse 2], [Chorus], [Bridge], [Final Chorus] — mark each part with a tag in square brackets, and after a hyphen in the tag say what happens in it.
Requirements: even meter in the verses, simple rhymes, short lines, a 4-line chorus that's easy to remember. Spell out numbers, avoid words whose pronunciation depends on their meaning. Keep the total under 2,000 characters.

In GPTunneL both steps share one balance: the lyrics in ChatGPT or another text model, the song in Suno.

What changed in Suno v6

Suno released v6 on September 9, 2026 and retired all its previous models at once. The family has three versions: v6 is the main one, v6-mini is the fast, lighter one, and v6-wild is the experimental one, with less predictable results. GPTunneL has v6 and v6 mini, and every track in this article was made on v6.

The headline features of v6 in Suno itself are editing a finished song in plain language and generating not just from text but from audio, an image or a video. For prompts, something else matters more: v6 was trained from scratch on a new dataset, so the same prompt may sound different on it than it did on v5 — it's worth re-testing your favorite phrasings.

FAQ

What language should I write a Suno prompt in?

The style can be written in English or another language — this article includes tracks made both ways. For a song in a language other than English, the combination "style in English, lyrics in the song's language" worked well in our tests. State the language of the song itself explicitly: "English lyrics" or "lyrics in Spanish".

How do I make a track with no vocals in Suno?

In Simple mode, write "no vocals" or "instrumental" in the description. In Advanced mode, leave only [Instrumental] in the lyrics field and add instrumental to the style. Both approaches were tested on tracks in this article.

Why does Suno sing different lyrics from mine?

In Simple mode, Suno writes the lyrics itself. If you need your own words, paste them into Advanced mode or straight into the description with "Sing exactly these lyrics", as in the example from the metatags section.

How much does a Suno generation cost in GPTunneL?

$0.24 per generation, which includes two tracks. There's no subscription: money comes off your shared balance only for finished generations.

Can I set the length of a track?

Roughly, yes: a jingle asked for "about 30 seconds" came out at 28 and 30 seconds. For a song, the amount of lyrics matters more: four lines gave about a minute, and the 1,488 characters of lyrics in "The Dawn That Stays" gave 4 minutes 37 seconds.

Try your own prompt

The best test of the formula is your own song: take a prompt from the article, swap in your topic and launch Suno in GPTunneL. No VPN needed, no subscriptions — and you can write the lyrics in a chat model on the same balance.