Skip to content

Foggy Hearts

Permanent short link: zylelabs.dev/foggy

A Cover workflow in Identity, Behaviour and Confidence.

Suno model: v5.5

Back to the guide · All song breakdowns

Open Foggy Hearts by Niri on Suno

Direct Suno embed of the original Foggy Hearts by Niri.

When I heard the original Foggy Hearts one thing really stood out to me, how clean the mix is. There are very few layers, the mix is well spaced out, the bass is behaved and the vocals are extremely clear. Experience has shown me this is the perfect recipe for a cover. It gives the model a clear structure to lock on to, with plenty of room to explore different styles. I wanted to keep the emotion in the vocals while giving the track some more momentum in the groove.

Foggy Hearts workflow: the original House song acts as an Anchor for the model, then three Cover stages establish musical Identity, refine Behaviour, and increase Confidence for the final Liquid Drum and Bass release.

Establish the new musical identity, define how that language should perform, then decide how boldly the model should reinterpret the source.

Stage Responsibility What changed here
Identity Define the musical language The style prompt established Liquid Drum and Bass
Behaviour Define how that language should perform Rhythm, Vocal and Priority instructions were refined
Confidence Decide how boldly the model should reinterpret the source Weirdness, Style Influence and Audio Influence changed

Niri’s original song used this prompt:

Deep Uk House, 124 Bpm, A♯ Minor, Low-Register Piano Stabs, Rubbery Donk Bassline, Sidechained Kick, Clipped Clap Pattern, Shuffle Hi-Hats, Distorted Diva Vocals, Light Autotune, Plate Reverb, Tape Saturation, Vocal Hook Repeats, Peak-Time Bounce, Bittersweet Longing, Early 2000S Club Mix

That prompt describes much more than a genre. It carries rhythm, instrumental palette, vocal character, production treatment, arrangement behaviour and emotional tone. The audio of the song carried those relationships into the Cover as an Anchor.

I know the destination, but not much else.

The first prompt I crafted was to set the tone and answer to me, how does the model react to these styles, and what does it produce:

Genre | Liquid Drum and Bass
Tempo | 174 BPM
Intro | Atmospheric | Liquid Flow
Bass | Bouncing Warm Reese | FM Growl | Sub Bass
Melody | Rhodes Stabs | Call and Response
Harmony | Cinematic Synth | Suspended Chords
Textures | Vinyl Dust | Evolving Pads | Granular Wash
Drums | Rolling Amen Variations | Ghost Notes | Tight Breaks
Vocals | Female Lead | Intimate Verses | Anthemic Attitude
FX | Reverse Swells | Tape Stops | Filter Throws
Dynamics | Long Crescendos | Sudden Silence
Priority | Groove
Motion | Constant Evolution
Mix | Wide Stereo | Controlled Sidechain | Warm Saturation
Mood | Motion through Hope
Structure | Constant Flow | Double Drop | Reflective Outro
Influence | 00's liquid | Modern Liquid
Control Setting
Weirdness 25%
Style Influence 70%
Audio Influence 60%

Open the Identity MP3 directly

The result showed that it had no issues making a Liquid Drum and Bass track out of the song, but there were some clear issues. The vocals sped up to match the tempo too much and lost the natural feeling the original had. The vocal vibrato felt extremely synthetic. That was the first thing I needed to fix, making the vocals fit the halftime synths. The other thing I noticed was that the drums felt way too relaxed for a Liquid DnB track, and the bass carried a lot of the house vibe into the cover.

This second pass addressed those problems directly.

What I heard Behaviour change Responsibility
The drums relaxed Added Rhythm, Rolling Liquid Breaks & Full-Time Drum and Bass Get the drums to lock onto that Liquid vibe
The arrangement became too static Added Driving Two-step, Continuous Forward Motion & Never Halftime to Priority while retaining Motion & Constant Evolution Reinforce motion as a non-negotiable behaviour
The vocal became more synthetic Replaced Intimate Verses & Anthemic Attitude with Relaxed Vocal Delivery & Natural Phrasing Ask for a less stylised, more human delivery

What I wanted to hear here was the liquid flow in the drums, the voice to return to that natural tone of the original song and the bass to push towards more forwards motion than the clubby bounce of the house original.

Highlighted text was added or revised in the Behaviour prompt.


Genre | Liquid Drum and Bass

Tempo | 174 BPM

Intro | Atmospheric | Liquid Flow

Bass | Bouncing Warm Reese | FM Growl | Sub Bass

Melody | Rhodes Stabs | Call and Response

Harmony | Cinematic Synth | Suspended Chords

Rhythm | Rolling Liquid Breaks | Full-Time Drum and Bass

Textures | Vinyl Dust | Evolving Pads | Granular Wash

Drums | Rolling Amen Variations | Ghost Notes | Tight Breaks

Vocals | Female Lead | Relaxed Vocal Delivery | Natural Phrasing

FX | Reverse Swells | Tape Stops | Filter Throws

Dynamics | Long Crescendos | Sudden Silence

Priority | Driving Two-Step | Continuous Forward Motion | Never Halftime

Motion | Constant Evolution

Mix | Wide Stereo | Controlled Sidechain | Warm Saturation

Mood | Motion through Hope

Structure | Constant Flow | Double Drop | Reflective Outro

Influence | 00's liquid | Modern Liquid

The controls remained unchanged:

Control Setting
Weirdness 25%
Style Influence 70%
Audio Influence 60%

Open the Behaviour MP3 directly

Holding the controls steady matters. It makes the Identity-to-Behaviour comparison easier to understand: one Rhythm field was added, while Vocal and Priority were rewritten.

The final release kept the Behaviour prompt unchanged. Only the influence controls moved:

Control Behaviour Final release Change
Weirdness 25% 30% +5 points
Style Influence 70% 80% +10 points
Audio Influence 60% 50% −10 points

I liked what the model was doing, I just needed to change how much here. I wanted it to drop the clubby vibe in the bassline so I reduced the Audio Influence a bit, then I wanted to push my prompt into the model a little harder, so I bumped up Style Influence. Weirdness was increased slightly to allow the model to explore a little more.

These are not magic settings, just what I was trying to do and how I attempted to get there.

Open Foggy Hearts by Zyle on Suno

The final prompt is not dramatically longer than the first prompt I started with. Foggy Hearts is not an argument for prompt inflation.

Successful prompt ≠ more words
Successful prompt = clearer responsibilities

The refinement makes particular fields accountable for particular behaviour:

  • Rhythm defines the full-time Liquid Drum and Bass pulse.
  • Priority protects forward motion and prevents halftime drift.
  • Vocals asks for relaxed, natural phrasing.
  • Motion continues to request evolution across the arrangement.
  • Influence controls decide how boldly the Cover moves away from its audio source.

That is the lesson I take from Foggy Hearts. The first prompt knew the destination. The successful revisions made the responsibilities clearer, and told the model how much to use my prompt vs the original audio.