Current for v5.5 · one toggle, four extra controls

Suno AI
custom mode

Same model, same credits, same speed - the only thing that changes is how much of the song you decide instead of Suno. Here's every field, every slider behind Advanced Options, four setups you can paste in, and the cases where Simple is honestly the better call.

1,000chars of style
5,000chars of lyrics
4advanced controls
0extra credits
The short answer

What custom mode changes

There's a toggle at the top of the Create page. Left is Simple, right is Custom. Flip it and one description box becomes three separate fields - Style of Music, Lyrics and Title - plus an Advanced Options panel with four controls Simple never shows you.

THE MISUNDERSTANDING

Simple mode isn't a weaker model. It's the same model, given fewer instructions.

Custom mode doesn't unlock better audio, more credits or a smarter engine. Both modes run the same v5.5 and cost the same 10 credits a generation.

What actually changes is authorship. In Simple mode you describe a song and Suno writes the words for you. In Custom mode you supply the words, and Suno performs them. That's the whole difference - and it's why Custom is essential for some jobs and pointless for others.

Simple

One box. Describe it and go.

  • A single song description field
  • Suno writes the lyrics from your description
  • An Instrumental toggle if you want no words
  • No section control - it picks the structure
  • No Advanced Options panel

Best for: exploring, finding a sound, birthday songs, anything where you don't care what the words are - and for genuinely fast ideas, where typing lyrics would slow you down.

Custom

Three fields plus the advanced drawer.

  • Style of Music - the sound, separately
  • Lyrics - your words, with section tags
  • Title - names the track, and nudges the mood
  • Full control of structure via meta tags
  • Advanced Options: Exclude Styles, Weirdness, Style Influence, Audio Influence

Best for: anything you'll release, anything with words that have to be those words, and any song that needs a specific shape rather than whatever the model reaches for.

The unpopular opinion

When Simple mode is the right answer

Nearly every guide tells you Custom is always better. It isn't - and reaching for it reflexively is how people spend forty minutes writing lyrics for a song whose sound they hadn't settled yet.

You're still hunting for the sound

Finding a genre takes ten generations, and writing lyrics for nine you'll throw away is wasted work. Lock the sound in Simple first, then move to Custom to write into it.

The words genuinely don't matter

Background music, a podcast bed, a shop playlist, something instrumental-adjacent. Suno's lyrics are perfectly serviceable when nobody is listening to them closely.

You want to be surprised

Custom mode gives you exactly what you asked for, which caps the result at your own imagination. Simple occasionally hands back a line you'd never have written - and that's a real reason to use it.

The workflow most people land on: explore in Simple until a generation makes you sit up, copy the style tags it used, switch to Custom, paste them into Style of Music, and write your lyrics into a sound you already know works.
The three boxes

What goes in each field

Getting this wrong is the most common Custom mode failure, and it announces itself immediately: anything in the Lyrics box that isn't a bracketed tag gets sung out loud. People have released tracks where a voice earnestly sings "upbeat pop, female vocals, 120 BPM".

Field oneStyle of Music1,000 chars

The entire sound world of the song: genre and subgenre, mood, vocal character, instruments, production and tempo, in roughly that order - Suno weighs the earliest tags most heavily.

Aim for 8 to 15 tags. Under five leaves the arrangement to the model's defaults; past twenty they start contradicting each other. Full method on the prompt guide.

Field twoLyrics5,000 chars

The words to be performed, plus section tags in square brackets - [Verse], [Chorus], [Bridge] - each on its own line. Tags organise the arrangement and aren't sung.

The box holds 5,000 characters but filling it is a trap. Past about 3,000 - roughly 40 to 60 lines - the model starts rushing to fit everything into the runtime and the phrasing goes flat.

Field threeTitle100 chars

Names the track in your Library and on the cover art. It's the field everyone ignores, and it does more than you'd think - the title feeds into the generation, so "Concrete Sunday" and "Untitled 4" don't produce identical songs.

Leave it blank and Suno names it for you, usually from a line in your chorus.

Character limits by model

The limits follow your model version, not your plan. A free account on v4.5 gets the same 1,000-character style box as a Premier account on v5.5.

Fieldv4.5, v5, v5.5v4 and olderPractical ceiling
Style of Music1,000200150-300 chars is the sweet spot - 8 to 15 tags.
Lyrics5,0003,000Stay under ~3,000 or phrasing gets rushed.
Title10080Short. It's a name, not a description.
Having 1,000 characters doesn't mean using them. The style box rewards precision, not volume - five specific tags beat fifteen vague ones, and a full thousand characters of description reliably produces mud.
The drawer nobody opens

Advanced Options

Four controls that only exist in Custom mode, and they're blunter and more reliable than any wording you can put in the style box. Using them also frees up characters for the things prompts are actually good at.

Exclude Stylesthe negative prompt that actually works

Typing "no drums" into the style box mostly teaches the model the word drums. This field is the documented control. Keep it to your two or three most important exclusions - a long list confuses the model as surely as a long prompt does.

Vocals creeping into an instrumentalvocals, singing, choir, vocal chops
Everything drifting toward EDMEDM, build-up, drop, sidechain
Weirdnesshow far from the genre's centre of gravity

Low is textbook and commercial. High deliberately fragments. 40-60% is the right default for an original song - past about 80% the model breaks things on purpose, which is useful for sample fodder and useless for a full track.

Safe & predictable
Sweet spot
Experimental
Glitch
0-20%20-60%60-80%80-100%
Style Influencehow literally your tags are obeyed

Low treats your tags as hints, high as instructions. Pair a high setting with a lean 5-8 tag prompt - a crowded tag list at maximum adherence gives you a mix where everything you asked for is present and nothing is audible.

Loose hints
Balanced
Strict
0-30%40-70%70-100%
Audio Influenceonly appears once you've attached a clip

How hard the result tracks an uploaded or recorded audio input. Around 55% keeps the melody while re-rendering everything around it. It pulls against Style Influence - move one at a time, or you'll learn nothing from the result. More on audio inputs.

Distant nod
Recognisable
Clip dominates
0-30%40-70%70-100%

The other controls sitting alongside

Persona or Voice

Attach a Persona and Suno auto-fills the style box from the source song - delete anything in there that contradicts it, or the identity drifts. Attach a Voice and you can drop the vocal-timbre tags entirely, since the singer is already decided.

Duration slider

Added on 20 July 2026, and web only - it never came to the apps. If you need a specific song length, you need a browser. Asking for one in prose does nothing in either mode.

Model selector

Pick the version that generates. v5.5 is current and the most expressive, but it also punishes vague prompts harder than v5 did - a thin prompt sounds more generic on it, not less. Custom mode is where that matters most.

Mobile gets a slightly different set. Since May 2026 the apps have a Vocal Gender selector, a Lyrics Model picker and Memory (which keeps your last prompt loaded). What they don't have is the Duration slider. Details on the Android app page.
Copy and paste

Four complete Custom mode setups

Each one is a full configuration - style box, lyrics scaffold and the settings that go with it. Paste them in, change the words, and you have a working starting point rather than a blank form. Click the icon on any block to copy it.

1 · A proper song, start to finish

The default setup. Verse, chorus, bridge, and a hard stop at the end.

Style of Musicindie folk, warm and nostalgic, soft female vocal with breathy delivery, fingerpicked acoustic guitar, upright bass, brushed drums, lo-fi tape warmth, 95 BPM
Lyrics scaffold[Intro] [Verse 1] four to eight lines, keep the syllable count even so the meter holds together across the verse [Pre-Chorus] two lines that lift into the hook [Chorus] two to four lines, the thing people remember [Verse 2] same shape as verse one, new detail [Chorus] repeat the chorus word for word - identical lyrics are the strongest signal the melody should return [Bridge] the contrast section, do something different here [Chorus] [Outro] [End]
Weirdness 50%Style Influence 60%Exclude noneTitle write one

2 · Instrumental that stays instrumental

For video beds, podcast music and anything a voice would ruin. Three defences, because one isn't enough.

Style of Musiclofi hip hop, calm and slightly melancholy, no vocals, dusty jazz piano loop, soft boom bap drums, warm sub bass, vinyl crackle, gentle tape wobble, 78 BPM
Lyrics field[Intro] [Instrumental] [Instrumental Break] [Outro] [End]
Exclude Stylesvocals, singing, choir, vocal chops
Instrumental toggle onWeirdness 30%Style Influence 70%

3 · Same artist, every time

Building a catalogue that sounds like one act rather than twelve. Attach a Persona, then get out of its way.

Style of Music - after attaching a Personarestless and melancholic, 138 BPM. Sparse verses, full band from the second chorus.
What to delete from the auto-filled style boxAny singer description, any genre that contradicts the source, any era or production note that fights it. The Persona already carries all of that - repeating it in different words is how the identity drifts.
Persona attachedStyle Influence 40%Weirdness 40%Keep the style box short

4 · Section-level direction

When the arrangement matters as much as the words. Pipe-stacked tags and scoped cues do what prose can't.

Lyrics with section control[Verse 1 - whispered, acoustic guitar only, no drums] the quiet opening, close-mic'd and bare [Pre-Chorus | Build | Strings Rise] tension for two lines [Chorus | Anthemic | Stacked Harmonies | Full Band] the payoff, and I MEAN it (harmonised) [Breakdown] [Beat Drops Out] [Final Chorus - gospel choir, full intensity] same words, bigger everything [End]
Style Influence 70%Weirdness 45%ALL CAPS for emphasisPipes beat commas
Course correction

Six Custom mode mistakes

All six are the default behaviour of someone who flipped the toggle without changing how they work.

01Style descriptions in the Lyrics box

The signature Custom mode error. Everything in that field is treated as performable text, so your genre notes get sung. Only bracketed tags are exempt. Genre, instruments, production and tempo all belong in Style of Music.

02Filling the style box because it's big now

Simple mode's small box enforced discipline. A thousand characters removes it, and people write paragraphs. The model doesn't read prose better than tags - it just gets more contradictions to reconcile. Stay at 8 to 15 tags.

03Writing 5,000 characters of lyrics

Suno has to fit your words into the runtime. Give it sixty lines and it rushes, flattening the phrasing that made the melody work. Cut to the best 40 lines and use Extend if the song genuinely needs more.

04Never opening Advanced Options

People spend twenty minutes rewording a prompt to stop a vocal appearing, when Exclude Styles does it in four words. If Suno gives you a dedicated control, the control wins - every time.

05Changing three things between generations

New style tags, a different Weirdness and a rewritten bridge, all at once. The result is worse and you've learned nothing about which change did it. One variable per generation - it's slower for two attempts and much faster by the fifth.

06Fighting the mode instead of switching back

If you've generated eight times in Custom and still don't like the sound, the lyrics aren't the problem. Go back to Simple, find the sound in three or four fast generations, then return with the style tags that worked.

Questions

Custom mode FAQ

What is Custom mode in Suno?

A toggle at the top of the Create page that splits the single description box into Style of Music, Lyrics and Title, and adds an Advanced Options panel. Same model, same credits - you just supply the words and the settings instead of Suno choosing them.

Does Custom mode cost more credits?

No. A generation is 10 credits either way, and returns two songs either way. The only thing Custom costs you is the time it takes to fill in the fields.

Is Custom mode available on the free plan?

Yes. It's an interface toggle, not a paid feature - free accounts get the same three fields and the same Advanced Options. What the plan changes is the model, WAV export and commercial rights, not the controls.

Why is Suno singing my style description?

It's in the Lyrics box. Everything there that isn't inside square brackets is treated as words to perform. Move genre, instrument and production notes to Style of Music.

What are the character limits?

On v4.5 and up: 1,000 for Style, 5,000 for Lyrics, 100 for Title. On v4 and older it's 200 / 3,000 / 80. The limit follows your model version, not your plan.

Can I make an instrumental in Custom mode?

Yes, and it's the more reliable way. Use the Instrumental toggle, put no vocals in the style box, [Instrumental] in the lyrics field, and vocals, singing, choir in Exclude Styles. Three defences, because any one alone occasionally fails.

Is Custom mode always better than Simple?

No. Simple is genuinely better when you're still hunting for a sound, when the words don't matter, or when you want to be surprised. Custom is better for anything you intend to release. Most people should use both, in that order.

Do I have to write my own lyrics?

No - the Magic Wand will draft them from a short idea, and it costs nothing. Generate them, then edit into the shape you want. Custom mode means you control the lyrics, not that you have to originate every line.

Where is the Duration slider?

Under Advanced Options, and web only - it arrived on 20 July 2026 and never came to the mobile apps. For an exact song length you need a desktop browser.

What should Weirdness and Style Influence be?

Weirdness 40-60% for most original songs. Style Influence 40-70%, and push it to 70%+ only when your tag list is lean. Change one at a time or you can't tell which did what.

Does Custom mode work on mobile?

Yes - the toggle is at the top of Create in both apps. Mobile adds a Vocal Gender selector and Memory, and is missing the Duration slider. Everything else matches the website.

Can I switch modes mid-project?

Freely - the toggle doesn't discard anything you've generated. The usual move is exploring in Simple, then flipping to Custom once a generation sounds right and building the finished version from there.

Find the sound first, then write into it

Custom mode is where songs get finished, not where they get discovered. Three fast generations in Simple, then flip the toggle and do the real work with a sound you already trust.