Suno AI Prompt Builder
A structured prompt builder for Suno AI music generation. Pick instruments, song sections, and transitions — get a copy-ready prompt optimized for Suno's Lyrics field.
Suno is an AI music generator that turns text prompts into full songs — vocals, instruments, mixing — in seconds. The quality of your output depends almost entirely on how well you write the prompt.
This builder assembles Suno-ready prompts deterministically from structured selections. No generation — every word is assembled from your choices.
Builder
Pick instruments and genre, select song structure, get a full-song prompt in one shot. Best for quick experiments.
[instrumental] [Intro – set the mood, low energy] no drums sparse (1–2 elements), leave gaps legato transition: gradual [Verse – establish main theme, medium energy] moderate (3–5 elements), leave gaps legato transition: build and release [Chorus – peak energy, full density] dense (full stack) accented, sustained transition: gradual [Outro – wind down, low energy] drums removed, bass fading out sparse (1–2 elements), progressive dropout sustained transition: fade in/out
Configure one section at a time across 8 control axes. Run once per section, stack the outputs. Best when you want precise control over each part of the song.
1 · Section
2 · Instrument Roles
3 · Rhythm & Groove
4 · Melody System
Constraints
5 · Density
Space rules
6 · Articulation
7 · Timbre
8 · Arrangement Logic
[Development – increase complexity, medium energy], piano as lead, bass as rhythmic ostinato, drums as accent stabs, straight groove, restrained with light variation drums, lead with counter melody in gaps, no overlapping melodic lines, moderate (3–5 elements), leave gaps, staccato articulation, warm acoustic, instruments take turns leading, transition: gradual
How to Use the Output
- Copy the Song Prompt from the builder above
- Open ChatGPT, paste the Metaprompt first, then paste your Song Prompt after it
- ChatGPT returns a polished Suno-ready version
- Go to Suno → paste the refined prompt into the Lyrics field
- Leave Style empty. Upload a Cover image if you want one — Suno ignores it for generation
Why the LLM refine step?
The builder outputs a structural sketch. A cheap LLM pass (GPT-4o mini is fine) adds specificity and smooths the language. The builder gives structure; ChatGPT gives voice.
Metaprompt
Copy this into ChatGPT before pasting your song prompt.
You are a Suno AI prompt specialist. Refine the following song prompt following these rules:
HOW SUNO READS PROMPTS:
- Paste into the Lyrics field — Suno reads this for both lyrics and arrangement
- Leave the Style field EMPTY — the prompt carries all style information
- Upload a cover image separately — Suno does not use it for generation
PROMPT RULES:
- [Section] headers are strong structural signals — keep them
- Be specific with instrument roles: "piano as staccato lead" >> "piano"
- Control density: "sparse", "dense", "leave gaps", "strip back after phrases"
- Use transition verbs: "sudden cut", "gradual", "fade out", "build and release"
- Energy curve: Intro=minimal → Verse=moderate → Chorus=peak → Break=stripped → Drop=punchy → Outro=fade
- Don't over-specify — Suno fills gaps creatively. Give structure, not sheet music.
REFINE THIS PROMPT (return ONLY the refined prompt, no explanation):
---
[paste your song prompt here]Example: Full Composition Prompt
A finished multi-section prompt. Stack one block per section and paste the whole thing into Suno's Lyrics field.
[Intro – minimal, sparse]
piano as lead, staccato chords, no drums, flute silent,
lots of space, leave gaps, sparse (1–2 elements),
transition: gradual
[Main Theme – light groove, medium energy]
flute as lead melody, piano as harmony chords, upright bass enters as rhythmic ostinato,
soft swing groove, drums minimal with light ride cymbal and no fills,
clear lead at all times, moderate (3–5 elements), leave gaps,
legato, warm acoustic,
transition: gradual
[Development – introduce counter melody, medium energy]
piano and flute alternating lead and counter melody, bass steady, drums restrained,
call and response, no overlapping melodic lines, avoid crowding, clear gaps between phrases,
moderate (3–5 elements), strip back after phrases,
transition: gradual
[Break – strip back, low energy]
drums drop out completely, bass minimal, flute as solo lead phrase, piano as accent stabs,
monophonic, spacious and exposed, sparse (1–2 elements), leave gaps,
sustained, warm acoustic,
transition: sudden cut
[Build – increase density, high energy]
strings enter as texture, rhythm tightens, drums gradually more active, layers stacking controlled,
no overlapping melodic lines, moderate to dense layering, no section at full density for long,
each section introduces one new element,
transition: build and release
[Drop – full energy release, peak energy]
synth lead as lead, brass as rhythmic accent stabs, strings as texture, drums fully active and punchy,
straight groove, busy syncopated percussion, tight synchronization, no melodic clutter,
dense (full stack), clean punchy,
transition: sudden cut
[Variation – controlled contrast, medium energy]
instruments take turns leading, alternate between full ensemble and reduced sections,
avoid all layers playing simultaneously, no section at full density for long,
remove elements before adding new ones,
transition: elements drop out
[Outro – wind down, low energy]
elements drop out progressively, piano and flute return as lead, drums fully removed,
slow and spacious, monophonic, sparse (1–2 elements), leave gaps,
sustained, warm acoustic,
transition: fade in/outHow Suno Actually Works
Most people treat Suno like a search engine — type something vague, hope for the best. Understanding what's happening under the hood makes prompting significantly more effective.
Suno's pipeline has 4 stages:
- LLM pass — your prompt goes to a language model that generates lyrics and structural tags (
[Verse],[Chorus],[Drop]) - Tokenization — lyrics and style tags are converted into abstract tokens encoding the vibe and flow
- Diffusion — a single mel-spectrogram is generated in one pass (not separate tracks per instrument — one unified "painting" of the whole mix)
- Vocoder — the spectrogram is decoded into audio via a neural vocoder (EnCodec / HiFi-GAN)
Why section headers matter: [Chorus] and [Drop] aren't just labels — they're strong energy signals baked into the model's training. The model has learned that [Drop] means full density, punchy, tight synchronization. You're not describing what you want; you're activating a pattern the model already knows.
Why the Lyrics field, not Style: The Lyrics field is what the LLM reads first. Style is a secondary conditioning signal. Putting everything in Lyrics gives you much more structural control.
The artifact problem: AI-generated music has a subtle telltale — spectral peaks from diffusion denoising steps. Hi-hats, reverb tails, and vocal sibilance are where it shows up most. Each model release reduces this, but it's not gone yet.
Want the full technical breakdown of how Suno generates music — latent space compression, why it doesn't sound like mud, and how they keep costs low enough to offer a free tier?
Hear It in Practice
Songs generated using prompts built with this tool: suno.com/@aislopp
Tips
- Instrument specificity beats instrument count.
strings as short staccato rhythmic ostinato>>strings - Density and space rules do most of the work.
leave gapsandstrip back after phraseschange the feel more than adding instruments. - Articulation beats timbre. Swap
sustained→staccatoon the same setup and compare outputs. - Arrangement meta-rules prevent monotony.
no section at full density for longalone fixes flat-sounding output. - The refine step is not optional for best results. The metaprompt teaches the LLM what Suno actually responds to.
[instrumental]at the top signals no vocals — use it if you want a pure instrumental piece.