[Prompting] For the curious
This is not a guide or advocacy for a particular style of prompting. I'm sharing this as a chat with chatgpt last night that I found interesting and is making me think about my prompting in different ways due to how v6 might work.
What do you think? Anyone already trying this way?
----------
Creative Prompting in Suno v6: What Changes from v5?
The move from Suno v5 to v6 changes something important about prompting.
The goal is no longer simply to describe what ingredients should be present in the music. Increasingly, the prompt can describe what the music should do.
That distinction matters.
Suno v5 already represented a major improvement in genre understanding, mixing, vocals and creative control. It was perfectly capable of handling descriptive prompts rather than simple keyword lists. But v6 pushes this further: Suno says the new model understands more of the language musicians use around vocals, instrumentation, structure, mood, references and overall feel, and can accept multiple directions within the same request.
For creative prompting, that encourages a different mindset.
From ingredients to relationships
A traditional prompt might look something like:
«dream pop, shoegaze, female vocal, synth bass, reverb, psychedelic, electronic, cinematic, 120 BPM»
There is nothing inherently wrong with that. It communicates several useful attributes efficiently.
But it mostly describes a collection of ingredients.
A more v6-oriented prompt might say:
«Nocturnal dream pop built around fragile guitar and intimate vocals. Begin hazy and restrained, gradually introduce increasingly synthetic percussion, then allow the final chorus to expand into something euphoric and overwhelming while preserving the melancholy underneath.»
The second version gives Suno something additional:
relationships between the ingredients.
The track has a starting state, a direction of travel, a transformation and an emotional constant.
That is where much of the creative potential of v6 lies.
---
The problem with giant tag clouds
Extreme prompting can sometimes produce fascinating results:
«outlaw country, samba jazz, riot grrrl, spiritual jazz, electronicore, reggae black metal, western swing, dream pop, Appalachian folk, polka dubstep, acid breaks, amapiano, synthwave, modal jazz...»
This can absolutely generate something unexpected.
But there is a difference between complexity and prompt entropy.
The problem is not that Suno cannot combine genres. The problem is that the prompt provides very little hierarchy.
Which genre is the foundation?
Which elements are rhythmic?
Which are production influences?
Does the song begin as country and transform into drum & bass?
Is the black metal influence supposed to affect the vocals, guitars or entire arrangement?
Should the dream-pop atmosphere remain throughout?
The generator has to answer those questions itself.
That can be desirable when randomness is the goal, but it reduces control.
A better approach is often to take the same strange ingredients and establish relationships between them:
«Appalachian outlaw soul driven by a fast broken-beat pulse. Weathered vocal and fingerpicked acoustic guitar remain at the centre while dream-pop pads and granular electronics gradually invade the arrangement. Midway, erupt briefly into riot-grrrl intensity before falling back into a haunting folk refrain.»
The prompt is still unusual.
But the weirdness now has architecture.
---
The central rule: maximise contrast, not clutter
Creative prompting does not require dozens of genres.
Often the strongest results come from establishing one recognisable musical identity and introducing two or three elements that should not naturally belong there.
For example:
«Psychedelic desert soul with broken-beat electronics, distorted fiddle and ghostly gospel harmonies.»
That already contains substantial creative tension.
The important question then becomes:
What happens between those elements?
Perhaps the fiddle remains traditional while the rhythm becomes increasingly electronic.
Perhaps the gospel harmonies appear only at the climax.
Perhaps the soul arrangement gradually disintegrates into granular noise.
Creativity comes from the relationship, not simply the number of labels.
---
A useful v6 prompt architecture
A strong creative v6 prompt can be built from seven components:
**Core identity
+ contrasting influences
+ performance character
+ sonic world
+ arrangement trajectory
+ major mutation
+ ending**
Not every prompt needs all seven, but this provides a useful framework.
---
- Give the track a musical spine
Start by establishing what the song fundamentally is.
For example:
«Psychedelic desert soul.»
«Nocturnal Appalachian dream pop.»
«Minimal chamber techno.»
«Dusty spiritual jazz with electronic percussion.»
This provides a centre of gravity.
Additional influences can then distort that identity without completely replacing it.
---
- Introduce a few unusual elements
Instead of adding another ten genres, choose a few distinctive elements.
For example:
«warped analog synthesizer, duduk and distorted fiddle»
or:
«Gnawa-style hand percussion, Japanese boogie bass and ghostly choir»
Specific combinations create character.
The goal is not maximum quantity.
It is maximum productive tension.
---
- Describe how the musicians perform
Performance language can be extraordinarily useful because it describes character rather than category.
Compare:
«acoustic guitar, drums, vocals»
with:
«nervous fingerpicked guitar, loose drums that drag slightly behind the beat, and a weathered vocal delivered almost like a confession»
The second description gives the track personality.
Useful performance concepts include:
fragile
hesitant
restrained
loose
aggressive
breathless
weathered
intimate
ritualistic
mechanical
imperfect
ecstatic
These words can sometimes contribute more than another genre label.
---
- Build a sonic world
One of the more interesting possibilities in v6 is the use of imagery.
Suno explicitly describes v6 as capable of creating from a “vibe” or feeling, rather than requiring everything to be expressed as conventional genre terminology.
That makes phrases such as these useful creative tools:
«submerged brass»
«dusty drums»
«flickering electronics»
«instruments appearing like memories»
«underwater machinery»
«a choir heard from another room»
«beautiful but slightly wrong»
«warm, human and decaying»
These aren't technical production specifications.
They are world-building instructions.
And sometimes that is exactly what a generative model needs.
---
Year and location: useful, but not magic
A year and place can be valuable additions to a creative prompt.
For example:
«Lagos, 1977»
or:
«Manchester, 1981»
or:
«Reykjavík, 1997»
These provide cultural and aesthetic context.
But they should be treated as shorthand, not precise engineering instructions.
“1978” is not one sound. A punk record, disco production, dub record and jazz recording made in 1978 could have almost nothing in common sonically.
So the stronger approach is:
«Lagos, 1977. Loose live rhythm section, warm bass, clipped guitar patterns, hand percussion, tape saturation and imperfect ensemble timing.»
Now the date and location establish the world while the rest of the prompt explains what matters about it.
For creative prompting, an especially interesting technique is to combine a historically grounded setting with something impossible:
«Lagos, 1977, gradually invaded by granular synthesis and futuristic digital percussion until the final section feels simultaneously 1977 and 2077.»
That creates deliberate tension rather than random genre accumulation.
---
BPM is still useful
BPM has not become pointless.
Tempo remains an important compositional instruction, particularly when rhythmic feel is central to the track.
The improvement is to combine BPM with behaviour.
Instead of:
«174 BPM»
try:
«174 BPM, fast and propulsive but spacious, with a vocal that feels almost half-time against the drums.»
Instead of:
«118 BPM»
try:
«118 BPM, patient motorik pulse with slightly lazy live drums.»
The number provides the approximate tempo.
The description tells Suno what that tempo should feel like.
If exact speed is unimportant, descriptive tempo language can sometimes be more creative:
«slow funeral procession»
«restless dance-floor pulse»
«loping midtempo groove»
«breathless percussion that feels as though it is constantly accelerating»
BPM should therefore be used when tempo is compositionally meaningful, not simply because every prompt is expected to contain a number.
---
Turn contradictions into arrangement
Contradictions are not necessarily a prompting mistake.
Unexplained contradictions are.
For example:
«minimal ambient, big room, tearout, acoustic folk»
gives the model several competing identities.
But:
«Begin as sparse acoustic folk with almost no percussion. Slowly introduce ambient electronics, then allow the final section to explode into an enormous distorted tearout rhythm.»
turns exactly the same contradiction into an arrangement.
This is one of the most important creative principles for v6:
If two ideas conflict, explain when or how the transition occurs.
The contradiction becomes part of the composition.
---
Think in transformations
A useful creative structure is:
recognisable state → intrusion → mutation → rebuild → unexpected ending
For example:
«Begin as intimate Appalachian folk.»
«Electronic percussion slowly enters.»
«Halfway through, the arrangement collapses into free jazz.»
«The original groove rebuilds using completely different instrumentation.»
«The final climax suddenly stops on a single dry acoustic sound.»
This gives the model a story to tell musically.
And importantly, the story does not have to be lyrical.
The arrangement itself can have a narrative.
---
One major surprise is often enough
Maximum creativity does not require constant chaos.
One genuinely strange event can give an otherwise coherent piece its identity.
Examples include:
«halfway through, collapse into a free-jazz section»
«remove every instrument except voice and bass for eight bars»
«replace the electronic rhythm with Appalachian strings»
«return to the chorus at half speed»
«let the final section gradually lose instruments until only room noise remains»
«suddenly change from synthetic production to the sound of a small live ensemble»
These moments work because the listener has first been given something stable enough to disrupt.
Surprise needs context.
---
Technical terminology should describe something audible
Terms such as:
granular synthesis
gated reverb
FM synthesis
dead-room drums
tape saturation
contact microphones
can certainly be useful.
But technical vocabulary works best when it serves an audible purpose.
Instead of:
«low bandwidth vocal»
consider:
«intimate vocal narrowed into an old-radio texture during the verses»
Instead of simply:
«mid-side processing»
consider:
«narrow centred vocal surrounded by extremely wide atmospheric synths»
The listener does not hear “mid-side processing.”
The listener hears the result.
Creative prompts therefore benefit from describing perception first and engineering second.
---
v6 and v6-wild should not necessarily be prompted the same way
Suno now explicitly separates v6 from v6-wild.
The company describes standard v6 as the more precise and controllable model, while v6-wild is intended to be more unpredictable, experimental and unconventional.
That suggests two slightly different creative strategies.
For standard v6:
Describe the strange idea clearly.
For v6-wild:
Describe a compelling premise and leave more decisions unresolved.
A v6 prompt might say:
«Haunted Appalachian soul fused with broken-beat electronics. Weathered vocal, fingerpicked guitar, distorted fiddle and unstable analog synths. Begin sparse and acoustic, gradually increase rhythmic complexity, collapse halfway into a short free-jazz section, then rebuild into an ecstatic gospel-influenced finale.»
A v6-wild version might simply say:
«Haunted Appalachian soul mutating into nocturnal broken-beat psychedelia. Fragile and human at first, increasingly unfamiliar as it develops. Include one major stylistic rupture and rebuild into something ecstatic but melancholy. Organic and futuristic, beautiful but slightly wrong.»
The second prompt deliberately leaves Suno to decide what the rupture actually is.
That uncertainty is part of the creative process.
---
The 1000-character principle
Long prompts are not inherently better.
A useful creative discipline is to fit the core idea into roughly 1000 characters.
That forces the prompt writer to prioritise.
Every sentence should ideally contribute one of four things:
identity
character
movement
surprise
For example:
«Midnight desert gospel fused with broken-beat electronic jazz and psychedelic folk. Start intimate: fingerpicked guitar, dusty upright bass, sparse hand percussion and a weathered half-sung vocal. Gradually introduce Gnawa-style rhythmic cycles, detuned electric piano, warped analog synths and ghostly harmonies. Let submerged brass and granular acoustic textures drift through the arrangement. Build into a euphoric chorus driven by elastic bass and syncopated drums, avoiding a conventional EDM drop. Midway, collapse into a brief free-jazz/Appalachian interlude, then rebuild piece by piece. End brighter and increasingly ecstatic with gospel vocals, distorted fiddle and polyrhythms before stopping suddenly on one dry acoustic sound. Beautiful, strange, human and futuristic.»
There are plenty of ideas in that prompt.
But they all describe the same musical journey.
That is complexity without entropy.
---
A reusable creative v6 template
A useful template is:
«[Core musical identity], with [2–3 contrasting influences]. [Performance character], featuring [signature instruments/textures]. [Optional year/location/recording world]. Around [BPM] with [description of rhythmic feel]. Begin [opening state], then gradually [transformation]. Midway, [major mutation or surprise]. Rebuild toward [final state]. End with [distinctive ending]. Emotional world: [image or feeling]. Avoid [unwanted cliché].»
The template is not intended to be filled mechanically.
Its purpose is to make sure the different instructions have relationships.
---
The bigger shift from v5 to v6
V5 improved Suno's ability to understand genre, mixing, vocals and creative direction. V6 does not make those techniques obsolete; it expands what can be communicated through natural language.
The biggest creative shift is therefore not:
“Stop using tags.”
It is:
“Stop expecting tags to do all the work.”
Tags remain useful for establishing genre, instrumentation, tempo and production characteristics.
But v6 gives prompt writers more reason to describe:
relationships instead of lists
movement instead of static attributes
performance instead of instrumentation alone
world-building instead of genre alone
transformation instead of contradiction
and intention instead of sheer quantity
Suno itself says that v6 can understand more complex musical directions and that “more direction” can produce better results.
The creative opportunity is therefore not to write the longest possible prompt.
It is to give the model a strong musical premise, describe the important relationships, establish a direction of travel and then leave just enough unresolved for something unexpected to happen.
In short:
v5 prompting could describe the ingredients of a strange song.
v6 prompting can increasingly describe how that strange song comes into existence.