Every track on this site started as a prompt before it was a song. Some of those prompts worked on the first try. A lot of them didn't — not because the idea was bad, but because the instructions Suno actually received didn't say what we meant. Here's what we've learned about closing that gap, and the tool we built ourselves to make it faster.
Two Boxes, Two Jobs
Suno splits your input into a Style/Description field and a Lyrics field, and they aren't interchangeable. The Style box is where genre, overall mood, era, and instrumentation palette belong — "moody synthwave, female vocal, cinematic" describes the whole track, once, up front. The Lyrics box is where structure and moment-to-moment direction live: `[Verse]`, `[Chorus]`, `[Whispered]`, `[Guitar solo]` — these are instructions about a specific point in the song, not a description of the song as a whole. Put a genre tag in the middle of your lyrics and Suno usually just tries to sing it; put a `[Chorus]` marker in the style box and it has nothing to attach to. Keep the two jobs separate and both boxes do what they're supposed to.
Bracket Tags Are Stage Directions, Not Lyrics
A bracketed tag in the Lyrics box — `[Bridge]`, `[Tape stop]`, `[Ad-libs]` — is a note to the band, not a line to sing. Suno reads it, applies it, and moves on. The formatting has to stay consistent for that to work: always `[Square Brackets]`, never `(Parentheses)` or ALL CAPS or a stray colon halfway through. Mixing styles is the single fastest way to confuse the model into treating a stage direction as an actual lyric — which is how you end up with a singer very earnestly performing the word "chorus."
One Instruction at a Time
It's tempting to stack everything you want into one line — `[Whispered, reverb wash, building tension, strings swell]` — but Suno tends to blur a pile of simultaneous instructions into mush rather than executing all four cleanly. One tag, on its own line, immediately above the section it should affect, reads far more reliably than four tags crammed together. If a moment genuinely needs more than one thing happening, it's usually better to let the first tag land for a line or two, then introduce the second.
Placement Is the Instruction
Where a tag sits matters as much as what it says. A tag placed directly above the line it's meant to change applies there; a tag left floating with nothing under it does very little. This is also why "wrong-place, right-tag" is such a common failure mode — the direction was correct, it just landed on the wrong bar.
Know the Categories, Then Stop Memorizing Them
Structure (`[Intro]`, `[Bridge]`, `[Drop]`), breakdown & FX (`[Tape stop]`, `[Vinyl crackle]`), vocal direction (`[Whispered]`, `[Rap verse]`, `[Choir]`), production notes (`[808 bass drop]`, `[Strings swell]`), instrumentation (`[Guitar solo]`, `[Sub-bass hum]`), mood & energy (`[Euphoric release]`, `[Haunting]`), and genre & style (`[Synthwave]`, `[Lo-fi hip-hop]`) — these are the categories that actually move a generation. You don't need to keep them all in your head. We built our own Suno Prompt Library for exactly this: click through the tags by category, watch the script assemble itself, then copy the whole thing straight into Suno's Lyrics box. It's the same tool we use before every release on this site.
Treat the First Generation as a Draft
Even a well-formed prompt is a starting point, not a guarantee. Generate, listen for which tags actually changed the output and which got ignored, then adjust and run it again. Over time you'll notice the same handful of tags reliably doing what you want for your sound — that shortlist, more than any universal rulebook, is the real payoff of testing.
A Compiled Example You Can Use
Everything above, put together. This one's built entirely from tags pulled off our own Prompt Library — copy the Style block into Suno's Style/Description field, copy the Lyrics block into the Lyrics field, and generate. Swap the words for your own; the structure is the part worth keeping.
Style / Description Box
Atmospheric alt-rock, warm male lead vocal, mid-tempo build into an anthemic chorus, subtle analog synth pads under driving distorted guitar, cinematic and emotional, modern indie-rock production polish.
Lyrics Box — "Ghost in the Faders"
[Intro]
[Dreamlike]
[Verse]
I type it out in brackets, hand it to the dark,
Something wakes up humming, chasing my rough spark.
No blood, no breath between us, just a line of code and time,
And somehow what comes back sounds more like mine than mine.
[Pre-Chorus]
[Building tension]
Four counts and the silence bends,
Something learns where the feeling ends—
[Chorus]
[Anthemic]
Ghost in the faders, wearing my name,
Turning the static into a flame,
You don't need blood to feel this way,
You just need me to say it plain.
[Post-Chorus]
Say it plain, say it plain...
[Verse]
Second take, wrong note, right room by accident,
I ask for thunder, you send back a quiet argument.
Every fix I whisper comes back changed, comes back new,
I don't know anymore where I end and you come through.
[Pre-Chorus]
[Building tension]
Four counts and the silence bends,
Something learns where the feeling ends—
[Chorus]
[Anthemic]
Ghost in the faders, wearing my name,
Turning the static into a flame,
You don't need blood to feel this way,
You just need me to say it plain.
[Bridge]
[Whispered]
Maybe the ghost was mine all along,
[Strings swell]
Just never had the wires to carry the song.
[Chorus]
[Key Change]
[Big Finish]
Ghost in the faders, wearing my name,
Turning the static into a flame,
You don't need blood to feel this way,
You just need me to say it plain —
Say it plain.
[Outro]
[Drums drop out]
...say it plain.
This example didn't stay hypothetical — we ran it. Hear the finished track below, on its track page, or read the story behind it.
