Suno Workflow: Simple, Custom, Then Meta Tags

Most “complete Suno guides” dump every menu item in order. You don’t need that on day one. You need one clean take, then a path to control structure without drowning in options.

Here’s the order that actually stuck for me: Simple until you hate the lyrics box being a black box → Custom when Styles and Lyrics must diverge → bracket tags only after the melody already works → section replace when one verse is wrong and the rest is fine.

Plans and credit numbers change; check the pricing page for commercial rights before you ship client work. Below is the workflow, not a rate card.

First ten minutes: Simple mode only

Open create. Stay on Simple. Write one sentence that names mood + language + voice gender + rough genre. Example:

Chinese Mandarin male vocal, mid-tempo indie pop, tired after work, clean guitar, not aggressive

Generate two clips. Keep the one where the vocal sits in the mix. Ignore cover art and titles until you have audio you wouldn’t mute.

If you wanted beds with no singing, flip Instrumental before generate—I covered that separately; this piece assumes you want a song with words.

When Simple stops being enough

Switch to Custom when any of these are true:

You already have lyrics (yours, a co-writer’s, or a draft from another tool).
You need Styles in English while the sung language is Chinese, Japanese, etc.
You keep getting the right vibe but wrong form (no chorus, endless verse).

Custom is three boxes that people mash into one prompt by habit. Keep them separate:

Lyrics — text the singer should roughly follow. Structure markers help later; plain paragraphs also work for a first pass.

Style of Music — English tags tend to stick: j-pop, bright synths, female vocal, 128 bpm, radio mix. One tag is fine; five is usually plenty. Dumping a paragraph of adjectives often makes the model chase everything and land nowhere.

Weird extras (names vary by UI version) — exclude styles you hate; upload a hum if the product supports audio-to-song; multi-voice / persona options when you actually need duets. Skip until the single-voice take works.

Copyright note: pasting someone else’s hit lyrics into Lyrics for “fun” is still a rights problem if you publish. Write originals or get clearance.

Open Suno in MidassAI

Lyrics without overthinking

Four sources I’ve used, none magical:

Write a short verse + chorus yourself (ugly drafts beat perfect blank pages).
Paste a poem or prose you own, then trim for singability.
Ask any chat model for a three-verse + chorus draft, then cut half the lines—AI lyrics run long.
Use Suno’s own lyric helper / Full Song if you want random structure; treat it as a sketch.

For Chinese topics, I still put language intent in Styles (mandarin vocal) and keep Lyrics in the language I want sung. Mixing random English filler into Mandarin lyrics usually produces accent soup.

Meta tags: after the take is good, not before

Square brackets in the lyrics field are stage directions, not poetry:

[Intro]
[Verse]
[Chorus]
[Bridge]
[Outro]

Official guidance across versions has been consistent on one point: a few tags beat a wall of them. Three structure tags often outperform ten instrument tags that fight the model’s arrangement instincts.

Voice ad-libs in parentheses—(yeah), (ooh)—can thicken a chorus. Overuse them and you get a track that never settles.

Instrument tags like [classical piano] can lock a mood; they can also freeze the mix into something thin. Use when the take is already close and you need a nudge.

Example block you can mutate:

[Verse]
streetlights blur on the ride home
I keep the radio low

[Chorus]
not today, not the office door
leave the badge in the drawer

Styles might stay: indie pop, male vocal, soft electric guitar, mid tempo, dry mix.

Editing without regenerating the whole song

Once you have a keeper:

Rename and cover art last—they don’t fix arrangement.
Extend when you need more runway for video; copy Styles so the second half doesn’t invent a new singer persona.
Regenerate when the concept is wrong, not when one word is wrong.
Song Editor → Replace Section (membership-gated on many accounts): keep the melody you like, swap one lyric block. That’s the tool for “chorus is gold, verse two is nonsense.”

If Replace Section isn’t on your plan, duplicate the Custom job, paste revised lyrics, and regenerate—then A/B the chorus against the original.

A compact “good enough” checklist

Goal	Mode	Focus
First draft today	Simple	One clear prompt, two takes
Locked lyrics + genre	Custom	Lyrics ≠ Styles languages if needed
Clear intro / chorus	Custom + tags	`[Verse]` / `[Chorus]` only
Fix one bad section	Editor	Replace Section or targeted regen
Bed with no vocals	Simple/Custom	Instrumental on (different article)

What I’d ignore early

Prompt-generator websites that spit 200-word Styles. Paying for analysis of a Spotify URL before you’ve made ten songs yourself. Obsessing over Pro vs Premier until you’ve burned a week of free/cheap credits learning what “good” sounds like for your ears.

Learn the loop: prompt → two listens → keep or kill → tighten Styles or Lyrics by one change. Meta tags and section replace are polish layers on top of that loop.