Suno Prompt Guide (v6): What 12,000 Real Prompts Taught Us

The short answer

A Suno prompt is four boxes in Custom mode, and each one has a job.

1. Style, up to 1,000 characters: the genre and era, the tempo, who is singing and how, the instruments, the production, the mood.
2. Lyrics, up to 5,000 characters: your words, with section tags like [Verse], [Chorus] and [Bridge] on their own lines.
3. Exclude Styles, up to 1,000 characters: what you do not want. This is where “no autotune” belongs, not in the Style box.
4. Title.

Here is one written the way we would write it, ready to paste into Style:

Late-90s neo-soul, 82 BPM, a warm female lead with a low, unhurried delivery and small runs on the last word of a line, Rhodes electric piano and a round upright bass, brushed drums sitting behind the beat, close and dry vocal, a little tape warmth, intimate and late-night

Everything below is what we learned about each box from 12,754 songs people made with HookGenius since April, plus our own tests inside Suno. The limits above were read off the live v6 composer on launch day (every Suno limit, measured).

How much do you want to leave to Suno?

This is the decision behind every prompt, and most guides skip it. Jesse Meria, who built HookGenius, puts it like this:

“How much do you want to allow Suno, the audio generator AI, to have creative freedom? … If you do not provide the context, Suno is going to fill the gaps with its own creative freedom.”

A three-word prompt is not wrong. It just hands Suno most of the decisions, starting with the voice and the arrangement. If that is what you want, keep it short and run it a few times. If you can hear the song in your head, write the parts you care about down, and Suno has less to guess.

His test for a finished prompt is whether you could read it and feel the song without hearing it:

“If you could not hear, you could read our articulation of a song, its structure, its style prompting, and its formatting from start to finish, and you would pretty much be able to feel what this song should sound like.”

Ten v6 prompts to paste

These are starting points for the genres people ask us for most: hip-hop, electronic, R&B, rock, pop, lo-fi, afrobeats, indie, country and Latin. Paste one into Style, change the voice and the mood to yours, and add your lyrics. More by genre in the hip-hop, EDM, R&B, rock and pop guides.

90s East Coast boom bap, 90 BPM, a male rapper with a calm, conversational flow that sits slightly behind the beat, dusty chopped soul sample, hard swung drums, deep upright-style bass, vinyl crackle, dry vocal up front, reflective

Dark modern trap, 140 BPM half-time, a melodic male rapper with clipped phrasing and ad-libs at line ends, heavy 808 sub-bass with slides, fast hi-hat rolls, eerie bell melody, atmospheric pads, wide polished mix, menacing

Deep house, 122 BPM, filtered female vocal chops and one soft sung hook, warm rolling bass synth, shuffled hats, Rhodes stabs, long reverb tails, builds over 16 bars, hypnotic and warm, late-night club

Modern R&B, 76 BPM, a silky female lead with breathy verses and stacked harmonies in the chorus, soft runs, warm sub bass, finger snaps, muted guitar, airy pads, intimate and confident

Alternative rock, 135 BPM, a raspy male lead, quiet verses that explode into a loud chorus, distorted guitar wall, driving bass, big room drums, raw and emotional, live-band feel

Bright 2020s pop, 118 BPM, a clear female lead with a light, playful verse and a full-voice chorus, punchy drums, plucked synth hook, sidechained pads, handclaps, big polished mix, euphoric

Lo-fi hip-hop instrumental, 75 BPM, Rhodes piano chords, tape-saturated drums with a lazy swing, warm analog bass, vinyl crackle and tape hiss, soft and nostalgic, study at night

Afrobeats, 104 BPM, a smooth male lead with a melodic, laid-back delivery and call-and-response backing vocals, log drum, shakers and congas, bright guitar riff, warm bass, sunny and danceable

Indie folk, 92 BPM, a soft male voice with a breathy, close delivery and a harmony in the chorus, fingerpicked acoustic guitar, upright bass, brushed snare, a little harmonica, warm room sound, intimate

Modern country, 98 BPM, a warm male lead with a slight twang and a storytelling verse, acoustic and electric guitars, pedal steel fills, steady kick and snare, bright chorus lift, heartfelt and hopeful

Reggaeton, 95 BPM, a rhythmic male lead singing in Spanish with a smooth, confident delivery, dembow beat, punchy 808, tropical synth plucks, shakers, polished club mix, high energy

What 12,754 real prompts taught us

1. Most people never describe the voice

In our 2026 data study of 3,483 tracks, three in four creators left the voice on the default. Only a quarter chose a singer at all, and the voice is what people complain about most. A voice line is the cheapest improvement you can make. Say who is singing and how they sing it, breathy or belted, conversational or behind the beat, and whether the vocal sits close and dry or wide and wet. More in Suno vocal prompts.

2. People write more than you’d think, and it works

The song ideas people typed to us had a median of 41 words, and nearly half ran past 50. Only 15% were under ten words. Long is fine if every phrase says something Suno can hear. The style prompts we send to Suno use a median of about 590 of the 1,000 characters. Use the room for the voice and the arrangement before you add more adjectives.

3. Half name an artist; describe the sound instead

47% of our songs started from an artist or a song someone named. Names are unreliable in the Style box. We have seen Suno read an ordinary word as a name: “capo”, as in a guitar capo, came back as a German rapper’s genre tags. So write what you hear in that artist instead: the era, the tempo, the voice, the instruments, how it is produced.

4. “No” in the Style box can backfire

When a style prompt says “no reverb” or “no autotune”, Suno does not reliably honour it, and sometimes renders the very thing. We measured it over 585 generations. Put everything you do not want in Exclude Styles, and keep the Style box to what you do want.

5. One line sung over and over can stop a render

In our tests on Suno v5.5, a lyric where one stock line made up most of what was sung was refused, and the same song with that one line changed went through. Length did not matter; the concentration did. Give your chorus a hook, and let the verses say new things.

6. One song in four uses a language besides English

24% of the songs with a language set included one other than English. If you mix languages, say which sections are in which, and write those lyrics in that language yourself. Suno sings what is in the Lyrics box.

Section tags that work in the Lyrics box

Put each tag on its own line above the lines it covers: [Intro], [Verse], [Pre-Chorus], [Chorus], [Bridge], [Outro]. Keep each section’s lines together, and write the chorus out in full each time it comes back. A short direction inside the tag, for example [Chorus | full band, voice opens up], is how we tell Suno what changes in a section. The full list is in every Suno meta tag.

Mistakes we see most

• Leaving the voice on the default.
• Writing “no X” in Style instead of Exclude Styles.
• An artist’s name instead of a description of their sound.
• A chorus that is one line repeated.
• Two instructions that fight each other, like “intimate” and “stadium” in one prompt.

Frequently asked questions

How we measured this

Counts come from 12,754 songs made in HookGenius by 2,702 people between April and October 2026. Our own testing accounts are excluded, and nothing anyone wrote is quoted. Field limits were read off the live Suno v6 composer on 9 September 2026. The negation and repeated-line findings come from our own controlled tests in Suno.