Skip to content

5. Building prompts

← Behavioural prompting · Guide contents · Next: Cover, Mashup and Generate →

Once I know the musical responsibilities, I write them one field per line:

Category | Descriptor | Descriptor | Descriptor

This is not special syntax that Suno requires. It is simply the format I use to keep the prompt readable. I can scan it quickly, compare two versions and see which line was meant to change the result.

Before I start filling fields, I try to finish one sentence:

I want to preserve [Anchor], place it inside [Destination], and make the result feel [Emotional direction].

That gives me the centre of the prompt. I can then decide what supplies the Kinetic Engine and whether the source and destination need a Bridge such as shared key, compatible phrasing or rhythmic force.

I usually build outwards in this order:

  1. Frame: genre, identity and influence.
  2. Musical material: bass, melody, harmony and synth voice.
  3. Engine: tempo, rhythm and drums.
  4. Surface: textures and sound design.
  5. Performance: vocals and articulation.
  6. Development: FX, dynamics and motion.
  7. Conflict resolution: priority.
  8. Finish: mix, mood and structure.

This is an order for thinking, not a form Suno expects. I leave fields out when they have no clear job. Filling every category for completeness usually gives me more language to interpret without giving the track a stronger direction.

Priority is where I decide what wins if the rest of the prompt pulls in two directions. It is not a list of everything I like.

Priority | Preserve Chip Timbres Throughout | Never Halftime

That line protects the chip identity if atmospheric textures start washing it out, and protects the full-time rhythm if the arrangement starts leaning into halftime. If I promote every idea to Priority, I have not resolved the conflict at all.

I try to make every revision answer something I heard. Foggy Hearts reached Liquid Drum and Bass on the first pass, so adding more Liquid Drum and Bass language would not have addressed its weaknesses. The problems were more specific:

What I heard Role responsible Revision
Drums drifting away from full-time motion Rhythm and Priority Add explicit rolling breaks, driving two-step and Never Halftime
Vocal phrasing becoming synthetic Vocals Replace attitude language with relaxed delivery and natural phrasing
Arrangement becoming static Priority and Motion Protect continuous forward motion and constant evolution

I held the influence controls steady while making those edits. That made the comparison mainly about Behaviour. Later, I kept the prompt unchanged and moved Weirdness, Style Influence and Audio Influence instead. That was a Confidence decision. Separating the two gave me a much clearer idea of what each pass had changed.

Some combinations are interesting because they pull against each other. I do not want to remove all of that tension. I want to decide where each side should be heard so the model is not left to resolve the entire argument on its own.

Tension A clearer assignment
Dream haze versus crisp breaks Put haze under Textures and crispness under Drums
Tape flutter versus crystal detail Assign them to different layers or sections
Sudden silence versus continuous motion Make the silence a brief transition rather than a structural stall
Equal genre influence versus a strong source Anchor Give the Anchor explicit priority instead of insisting on equality
Genres | [Primary genre]
Tempo | [BPM]
Bass | [Movement] | [Sub foundation]
Melody | [Anchor or hook]
Harmony | [Chord language]
Rhythm | [Groove]
Drums | [Break type] | [Detail]
Vocals | [Voice] | [Delivery]
Motion | [Evolution]
Priority | [Non-negotiable behaviour]
Mix | [Width] | [Colour]
Mood | [Emotional direction]
Structure | [Flow] | [Ending]

I use this as a starting shape, then remove anything that does not help with the track in front of me.


← Behavioural prompting · Guide contents · Next: Cover, Mashup and Generate →