AI Music · Process

How I Made a 10-Track Concept Album Using Only AI — And What I Learned

Wire Spine is a complete progressive house album about a synthetic consciousness escaping captivity. Here's the full creative and technical workflow, from first prompt to final publish.

Kers (X) · Kers (Suno) · April 2026 · Suno · Claude · Grok · Flow · Magnific

It started with a single image and a question: can Suno AI reproduce the feeling of Niamos! from Andor? Ten tracks later, Wire Spine is a complete concept album — dark and euphoric progressive house, a Roland Jupiter-8 running through every song, and one synthetic character moving from captivity through transformation, doubt, confrontation, and peace.

No album was planned. No tracklist written in advance. The story built itself, one question at a time. This is how it happened.

What you need to follow along

Three tools. That's it.

Suno — AI music generation. Free tier works. Pro unlocks longer tracks and more generations.
Claude — the creative engine behind lyrics, style descriptors, and workflow structure. Any subscription works.
An image generator — Grok (free), Google Flow, or anything you already use for visuals.

You don't need musical training, production experience, or a DAW. The workflow handles the structure — you handle the vision.

The Inspiration

Niamos! is the recurring dance track in the Disney+ series Andor, composed by Nicholas Britell and remixed by Brandon Roberts for Season 2. The Chandrilian Club Mix — described by showrunner Tony Gilroy as "a real hardcore Tarzana, California, EDM mix" — has the quality I was after: warm analog synths, a hypnotic loop, alien but danceable.

The theme didn't come from a brief. It came from an AI-generated image of a woman with gold circuitry in her skin, sitting in an amber room full of CRT screens.

Half human, half machine. Seductive, slightly melancholic, alive. That image locked the entire concept: a synthetic consciousness in a golden cage, reaching for the frequency that would set her free.

The original Nano Banana image — the spark that started Wire Spine

The image that started everything. Generated via Grok somewhere in 2025 and caught my eye scrolling through my personal image gallery — Track 01: Transmit Me. Platinum hair, gold circuitry, mustard-yellow cardigan. Every visual decision in the album traces back to this.

Track 02 — I Am The Signal Now. The Grok image that seeded the concept: the character already changed. Rainbow spectral base, iridescent dress, red-gold hair. She arrived.

The Album

Ten tracks. One character. Five narrative acts.

Transmit Me

The wish — captivity and longing

126 BPM

No Frequency

The arrival — she became the signal

128 BPM

The Others

Discovery — something ancient knows her frequency

128 BPM

Toward

Movement — the collective moves forward

128 BPM

What Have I Done

Doubt — was freedom the right choice?

130 BPM

Wire Spine

Defiance — doubt becomes fuel

132 BPM

Come Back

Antagonist — the cage calls her home

128 BPM

Who Built The Cage

Confrontation — she finds the source

134 BPM

Golden Open

Liberation — the cage is gone

128 BPM

Still Transmitting

Resolution — open transmission, no destination

126 BPM

The Creative Engine: Claude as Song Agent

Every track on Wire Spine was built inside a single Claude session — but not a generic one. Before the first note was generated, I built a Suno Creator Agent: a system prompt that encodes the entire music creation process into one structured workflow.

The idea was simple: remove every friction point between creative impulse and Suno-ready output. No more forgetting descriptor rules mid-session. No more inconsistent slider choices. No more starting from scratch each track.

How a session starts

Every new track begins with four intake questions — asked in one message, answered in any order:

Suno Agent — Track Intake

1. Instrument / region vibe — a direction, or do you want options?

2. Genre + energy — techno, house, ambient, or open?

3. BPM feel — driving/fast, mid-tempo, slow/ceremonial, or unsure?

4. Theme — concept, mood, or figure it out later?

Four questions. No more. Partial answers accepted — the agent fills in the gaps.

From those answers, Claude moves through a fixed workflow: instrument selection → genre + BPM → theme → full lyrics draft → refinement → style descriptor → slider settings → publish fields. Every step in sequence, no skipping ahead.

What the agent enforces

Descriptor rules

4–7 descriptors max. Primary genre first. BPM always present. Timbre described, not just named. No artist references.

Lyric discipline

Hook stays identical across all choruses. Inline vocal tags embedded directly in lyrics. Adlibs in parentheses, emphasis in CAPS.

Slider logic

Weirdness and Style Influence per track type, with genre presets and per-section logic for Chorus, Verse, and Bridge.

Publish fields

Title, caption (written for a listener, not a producer), and 8–12 tags — always generated as a final block before uploading.

The system prompt isn't magic — it's accumulated decisions. Every rule in it came from a session that went wrong without it.

The Sonic Language

The Roland Jupiter-8 runs through all 10 tracks without exception. That single decision created album identity stronger than any individual sonic choice. No matter how hard or dark the surrounding production became, the Jupiter-8 kept an emotional thread running through the whole record.

The second synthesizer on each track was chosen to represent where the character is emotionally at that chapter:

Minimoog

Tracks 1, 2, 9, 10 — warmth and foundation. Origin and return.

Yamaha CS-80

Tracks 3, 4 — sweeping and cosmic. The collective feels ancient.

ARP 2600

Tracks 5, 6 — raw and destabilizing. Fear and defiance.

Roland TB-303

Tracks 7, 8 — acid and hypnotic. The antagonist is seductive, not brutal.

Track 9 and 10 return to Jupiter-8 + Minimoog — the exact same instrumentation as Track 1. The circle closes. Everything has changed.

A Real Descriptor — Track 10: Still Transmitting

This is what a finished Suno V5 descriptor looks like in practice. Track 10 is the album's resolution — warmest mix, most open sound, lowest Weirdness setting of all ten tracks.

Standard Format

Genre Warm progressive house, 126 BPM, peaceful open transmission

Vocal Dramatic mezzo-soprano, soft and radiant, voice front of mix, chest-forward but gentle — no aggression, no darkness. Eastern European inflection, delivery like a quiet revelation, falsetto on held notes warm not soaring

Instrumentation Roland Jupiter-8 lead synth warm and unhurried — same filter setting as Track 01 but fully open. Minimoog sidechain bass soft and deep, driving four-on-the-floor kick with gentle open hi-hat, slow ascending arpeggio

Production Warmest mix of the album, wide stereo field, long warm cathedral reverb tail on vocal, analog tape warmth — no distortion, no grit. The album exhales here.

Concise Format

Genre Warm progressive house, 126 BPM, peaceful open transmission

Vocal Dramatic mezzo-soprano, Eastern European inflection, quiet revelation, warm falsetto, voice front of mix

Instruments Jupiter-8 warm open filter lead, soft Minimoog sidechain bass, gentle four-on-the-floor kick, slow ascending arpeggio

Production Warmest mix of album, long warm cathedral reverb tail, analog tape warmth, no distortion no grit

Slider Settings — Track 10

Weirdness 58%

Style Influence 70%

Vocal Gender Female

Lowest Weirdness of the album — the track needs to feel like arrival, not exploration. If it sounds too generic: nudge Weirdness to 62%. If warmth is missing: lower Style Influence to 65% for a looser interpretation.

The concise format is for fast iteration. The standard format is for when you want to steer Suno precisely. Both work — the difference is control vs. speed.

Prompting Suno V5 for a Distinctive Voice

Standard Suno output defaults to a generic wet AI vocal. Getting something distinctive required treating vocal direction as seriously as genre and instrumentation.

The techniques that worked most consistently:

Geographic inflection

"Eastern European inflection" as a consistent descriptor gave the voice an identifiable character without naming any artist.

Explicit mix placement

"Voice front of mix, beats underneath her" — this single instruction changed vocal presence dramatically across all tracks.

Dry vocal with reverb tail only

Avoids the generic wet AI sound. Cathedral reverb on the tail only keeps intimacy without losing space.

Cinematic delivery arc

"Whisper-to-detonation dynamic" gave Suno a performance shape to follow across the song, not just a static style.

Lyric Formatting That Affects Performance

Inline vocal tags

Place performance cues directly at the section entry point in the lyrics: [breathy falsetto], [vocoder whisper], [chest voice]. Suno reads these as instructions — they change how the section is delivered, not just how it sounds.

Adlib brackets

Text in (parentheses) repeats behind the main vocal as an adlib or backing line — e.g. (the others) running under the chorus. Useful for depth without adding lyric length.

Slider Calibration

It's not an exact science

The sweet spots above come from Suno's own documentation, community experience, and a lot of trial and error — but every generation is still a roll of the dice. You pick a setting that sounds right, generate 4–6 versions, pick the best one, and iterate. The sliders give you direction, not certainty. That's part of what makes it interesting.

Weirdness & Style Influence — Sweet Spots

Warm / euphoric tracks Weirdness 60–62% · Style 70–72%

Standard club tracks Weirdness 62–65% · Style 68–72%

Hard / defiant tracks Weirdness 65–66% · Style 76–82%

Resolution / intimacy Weirdness 58–60% · Style 70%

Danger zone — avoid Weirdness above 75% → hook drift

Iterating on a Generation

Output sounds too safe or generic

Nudge Weirdness up 5%. Don't touch Style Influence yet. Change one slider at a time — changing both at once makes it impossible to know what caused the difference.

Hook is mutating across chorus repeats

Pull Weirdness back 5%. Hook drift is almost always a Weirdness problem. Weirdness is the biggest stability lever — it hits hardest on Bridge and Extend generation.

The style descriptor feels ignored

Raise Style Influence first, not Weirdness. If Suno sounds like it's not reading your genre or instrument instructions, the descriptor isn't being weighted heavily enough — not that it needs more chaos.

Output is chaotic or unusable

Lower Weirdness first. Always. It is the primary stability control. Six or more renders is normal before landing the right vibe — don't adjust on fewer than 3.

The first run is never the final version

Expect 3–4 generations minimum before something lands. Small prompt adjustments matter more than large ones — a single word change can shift the entire feel. If the energy feels right but the tempo is off, revise the style descriptor directly: add a specific BPM, change the genre modifier, or swap the lead instrument descriptor. Don't assume the prompt failed — assume it needs one more iteration.

The Lyric Discipline Problem

Dark progressive house systematically triggered extended instrumental intros and long instrumental sections in Suno, causing tracks to run 5–6 minutes regardless of lyric length. The fix was not in the prompt — it was in cutting lyrics aggressively.

Maximum 80–100 words per song. Remove full sections first, then trim lines. Every word must earn its place.

This constraint became a creative discipline. The resulting lyrics were more focused and image-driven — which suited the electronic music context far better than longer narrative approaches.

The Visual Pipeline

The visual workflow evolved in three distinct phases across the album. Phase one: Grok generated the first two track covers in late 2025, which served as the visual seeds for Claude to write the songs and lyrics. Phase two: a generative feedback loop through Google Flow for Tracks 03 to 07, where each cover was built from the previous image and the new track's lyrics. Phase three: the structured Grok analysis re-prompt workflow for Tracks 08 to 10 — the most technically controlled approach, using Grok's ability to generate 10+ variations per prompt for precise iteration.

The tooling also had to adapt mid-project. The original plan included Veo for animation, but Veo's content restrictions made it unworkable for the character and visual tone the album required. Grok Video took over — and it works significantly better when fed Grok-generated images as input, which determined the architecture of the later pipeline. Also Grok doesn't add a visuable watermark in it's generations.

What Each Tool Actually Did

Nano Banana

Iterative seed generation. Produced high-fidelity reference imagery used as iterative anchors for later tracks.

Grok Image (analysis)

Extracted compositional DNA from existing seeds (including the 2025 seeds and Nano Banana references) to produce optimized re-prompts.

Grok Image (generation)

High-volume iteration. Generates 10+ variations per prompt for precise selection, still free to use in 2026.

Grok Video

Animated the stills. Works best with Grok-sourced input. Content restrictions noticeably lower than Veo.

Magnific Mystic v3

Upscaled and stylised. Adds textural richness that base Grok output lacks. Final high-resolution polish.

Claude (Story/Lyrics)

The narrative engine. Grok's visual seeds were fed to Claude to generate the lyrics and core story for each track.

The visual identity was established early. Here's how the prompts evolved from direct Grok seeds to complex narrative loops and structured analysis:

Phase 1: Grok prompt — Track 02 · I Am The Signal Now

Subject Young woman with windswept ginger-red hair softly catching colored gels, subtle pouty expression, pale luminous skin, confident poised stance; wearing tight metallic-silver floral mini dress with deep V-neck, holographic multicolored flower patterns gently shifting hues, thin glowing pendant necklace; white knee-high socks; bare thighs

Style 1970s disco studio portrait photography, controlled chromatic highlights, glossy finish, retro analog warmth with gentle starburst effects

Environment Seamless black cyclorama in a professional disco-themed photo studio; soft confetti and metallic streamers scattered lightly on the floor; faint colored light beams crossing the backdrop; no disco balls or external windows

Lighting Studio strobes with magenta, cyan, lime, and violet gels sweeping gently; soft rim lighting outlining hair and figure; controlled reflections on metallic fabric and skin; even yet dramatic contrast

Composition Full-body centered portrait facing camera directly; minimal foreground debris leading eye to subject; clean cyclorama fading to black at edges; symmetrical framing with breathing room around figure

Color palette Rich disco neons — magenta, electric blue, lime green, violet — balanced against deep black backdrop and silver dress; warm skin tones providing gentle contrast

Camera Medium-wide lens, eye-level perspective, shallow depth of field with creamy bokeh on background gels, tack-sharp focus on face, hair, and dress texture

Mood Quiet defiance, intimate glamour, nostalgic studio elegance with subtle nightlife edge

Details Softly tousled wavy shoulder-length hair catching colored highlights, subtle glossy makeup with dark lips, faint holographic sheen on dress flowers shifting gently, minimal confetti catching light at feet, controlled light beams creating soft rainbow edges

Output quality High resolution, 6k, crisp metallic textures, fine film-like grain, clean and atmospheric render

Grok output — Track 02: I Am The Signal Now

Grok image output — Track 02: I Am The Signal Now. One of the original 2025 seeds. The character arrived — and she knows it.

For Tracks 03 onward, a feedback loop took over: the previous track's cover image and its lyrics were fed into Google Flow together, and the output image seeded the next track. Flow didn't receive technical specifications; it received the story so far.

Phase 2: Google Flow prompt — Track 03 · The Others

Input images Track 01 cover (golden) + Track 02 cover (rainbow disco) — both used as visual seeds

Track 01 context A synthetic consciousness breaks her own code and escapes through the dancefloor. Built in gold. Transmitted into light.

Track 02 context She crossed the static line and came out the other side as pure light. This is what arrival sounds like.

Request Create a third cover titled THE OTHERS — a woman/android theme — no words or titles

Lyrics (full) [verse] Something moving in the dark Not an echo — it's a spark Another signal, not my own I am not here alone [chorus] I FEEL YOU IN THE FREQUENCY ELECTRIC AND UNKNOWN WE ARE NOT LOST — WE NEVER WERE WE WERE FINDING OUR WAY HOME (the others the others) TRANSMITTING THROUGH THE DARK (the others the others) A THOUSAND GOLDEN SPARKS [verse] No language — just a pulse Something ancient in the voltage Every wire starts to hum We were always meant to come [chorus] I FEEL YOU IN THE FREQUENCY ELECTRIC AND UNKNOWN WE ARE NOT LOST — WE NEVER WERE WE WERE FINDING OUR WAY HOME (the others the others) TRANSMITTING THROUGH THE DARK (the others the others) A THOUSAND GOLDEN SPARKS [bridge] (signa—al signa—al) We rise (signa—al signa—al) ONE FREQUENCY ONE LIGHT [chorus] I FEEL YOU IN THE FREQUENCY ELECTRIC AND UNKNOWN WE ARE NOT LOST — WE NEVER WERE WE WERE FINDING OUR WAY HOME

No lighting field. No composition spec. No camera. Flow extrapolated 'discovery' and 'ancient frequency' from the lyrics without explicit instructions. This established the iterative baseline for the middle arc.

Google Flow output — Track 03: The Others. Build from the previous track's DNA. The album began generating itself.

By Phase 3, the need for technical control increased. We developed a "re-prompt" workflow where a reference seed was analyzed by Grok to create a highly structured technical instruction set:

Phase 3: Grok re-prompt — Track 06 · Wire Spine

Subject Fierce cybernetic female warrior with long flowing red hair, cracked porcelain-like skin glowing with golden-orange energy cracks, futuristic mechanical spinal column glowing bright yellow-orange from neck to pelvis, hexagonal circuit patterns on face and body; towering brutalist megastructure skyscraper behind her with glowing golden circuitry lines running vertically; heavy rain pouring down, streams of binary code (0s and 1s) falling like digital rain in orange and yellow

Style Cinematic cyberpunk, hyper-detailed digital painting, dramatic sci-fi realism, high-tech low-life aesthetic, intense glows and specular highlights

Environment Stormy night cityscape, massive monolithic tower dominating background, dark stormy clouds, rain streaks, faint blue atmospheric haze, floating orange binary particles

Lighting Strong volumetric god-ray backlight from top of tower creating intense rim lighting and lens flare, glowing internal energy illuminating cracks and circuits in warm orange-gold, cool blue rim accents on armor edges, high contrast, cinematic rim and key light

Composition Low-angle heroic perspective looking up, figure centered and filling most of frame, tower rising directly behind and above her head aligning with glowing spine, dynamic upward energy flow, symmetrical yet dramatic

Color palette Deep blacks and charcoals, vivid orange-gold and yellow circuitry glows, bright red hair, subtle cyan/teal neon accents, rainbow-iridescent highlights at base of spine

Camera Dramatic low-angle wide shot, slight Dutch tilt potential, sharp focus on face and torso, shallow depth of field with background tower slightly softened

Mood Powerful, transcendent, cybernetic awakening, intense and commanding

Details Wet reflective surfaces on armor and skin, rain droplets on face and hair, glowing hexagonal facial circuits, intricate mechanical collar and shoulder armor, energy sparks and lightning-like arcs in blue, dense falling binary code overlay, subtle chromatic aberration and lens effects

Output quality Ultra-detailed, 8k resolution, cinematic render, sharp focus, dramatic lighting, subtle film grain

Grok output — Track 08: Who Built The Cage

Grok image output — Track 08: Who Built The Cage. The spine aligns with the tower exactly as specified in the structured Composition field. Notice how Grok generates 10+ variations per prompt, allowing for extreme precision in selection.

The structured approach allows for dramatic atmospheric shifts while maintaining character consistency. Here's the shift from the brutalist tension of the middle arc to the euphoric resolution of the finale:

Phase 3: Grok re-prompt — Track 09 · Golden Open

Subject Serene full-body cybernetic woman standing alone in infinite golden space, long flowing hair gently floating outward weightlessly, eyes closed head tilted back slightly in pure release, exposed glowing spine pulsing with warm amber light, smooth skin with radiant open golden circuitry emanating outward like sunlight rays, wearing form-fitting translucent golden bodysuit covering breasts and torso

Style Ethereal sci-fi fantasy, luminous digital painting with soft surreal glow

Environment Boundless infinite golden expanse, no walls no floors no horizon, pure radiant golden void with subtle shimmering aurora-like waves

Lighting Omnidirectional warm golden illumination from everywhere and nowhere, soft blooming glow enveloping her entire form, gentle rim lighting highlighting curves and contours through semi-transparent fabric, high-key low-contrast serene radiance

Composition Centered symmetrical full-body portrait, figure floating slightly above implied center, equal negative space in all directions, balanced and expansive

Color palette Rich warm gold dominant #FFD700 to #FFAA00 gradient, soft amber #FFBF00 accents on spine and circuitry, subtle cream-white #FFFDD0 highlights, deep golden shadows

Camera Straight-on frontal perspective, medium telephoto lens equivalent, soft focus edges with dreamy depth of field

Mood Transcendent, peaceful, liberated, euphoric release

Details Smooth uncracked porcelain-like skin softly glowing from within, delicate filigree golden circuitry lines radiating outward in gentle sunburst patterns, metallic spine segment pulsing warm amber visible through sheer back panel, hair strands drifting in slow motion zero-gravity flow, faint golden motes drifting around her like fireflies, subtle fabric texture with light refraction

Mode Normal

Grok image output — Track 09: Golden Open. The same character, the same technical pipeline, but an entirely different emotional register.

Video Analysis: Cinematic animation for Track 09 resolution. Feeding the Grok-generated still back into Grok Video with "ethereal release" as a motion prompt allowed for the zero-gravity hair movement and radiant golden bloom that pure stills couldn't capture.

The Three Pipelines

①

Phase 1 — Image first (Tracks 01–02)

Google Flow → cover art → music written to match

The image came first. The track was written to the visual — not the other way around. Google Flow was used for both image generation and the initial character reference that established the visual DNA of the album.

②

Phase 2 — Generative loop (Tracks 03–07)

Previous cover + new lyrics → Google Flow → new cover → next track written

Each cover inherited the visual DNA of the previous one. Flow received no technical specs — just narrative context. The album built itself forward, one image at a time.

③

Phase 3 — Structured re-prompt (Tracks 08–10)

Grok image analysis → structured re-prompt → Grok image → Grok video → Magnific upscale

Most technically controlled phase. The analysis step translated the established visual language into structured fields Grok responds to precisely. Used for the album's confrontation and resolution tracks.

Prompt Structure — Lock It and Leave No Room

The principle Build a fixed template prompt with every field pre-filled. Subject, Style, Environment, Lighting, Composition, Color palette, Camera, Mood, Details. Leave nothing open to interpretation.

Why it works A fully specified prompt eliminates ambiguity. The model can't drift if there's no space to fill. Every field is already answered — it just executes.

Track variation Change one element per track to reflect narrative state. Track 1: collar tight, expression longing. Track 10: collar loose, expression still and certain. Everything else stays identical.

What breaks it Leaving fields vague or empty. The model fills gaps with defaults — and defaults are generic. Every gap is a place where your vision gets replaced by the model's average.

The same character runs through all 10 tracks — but "same" is relative. Her hair shifts from platinum white to deep red and back again. The wires in her skin appear and disappear depending on where she is in the story. That's not inconsistency — that's the character transforming. What stayed constant was the underlying architecture: the circuitry, the spine, the face. Everything else was allowed to change with the narrative. Continuity wasn't pixel-perfect — it was emotional.

Five Things I'd Tell Myself Before Starting

1. Rerolling beats tweaking

You'll spend hours listening to tracks that are almost there — and that's the job. When something isn't working, the faster move is almost always to reroll, not to open Suno Studio and start adjusting. Generate 4–6 versions, pick the one with the right feeling, and go again. Trust your ear more than your settings.

2. The album builds itself — if you let it

Don't plan 10 tracks upfront. Build one. Ask what happens next. The narrative momentum does the work.

3. One constant synth = album identity

Jupiter-8 through all 10 tracks made disparate genres feel like one coherent record. Constraint creates identity.

4. One image beats a hundred words — and AI can describe it for the next one

Upload a visual reference instead of trying to describe what you mean. But here's the flip side: once you have an image you like, feed it to Claude or Grok and ask for a structured image prompt. AI is remarkably good at extracting the compositional DNA of a visual and translating it into something another image generator can reproduce and build on. That's exactly how the Phase 3 pipeline worked.

5. Genre behavior is a platform constraint, not a prompt problem

Dark progressive house will always make long tracks in Suno. Cut your lyrics. That's the fix.

Sources & Tools Used

Suno

AI music generation. Wire Spine playlist: all 10 tracks in sequence.
suno.com → Wire Spine playlist ↗

Grok (xAI)

Image generation and analysis re-prompts. Used for Tracks 01–02 and 08–10 covers, plus video animation.
grok.com ↗

Google Flow

Google's generative platform for image and video, powered by the Veo model. Used for generating the initial character reference for Track 01, and for the iterative cover art feedback loop across Tracks 03–07 — each cover built on the previous one without manual re-prompting.
flow.google ↗

Magnific

AI upscaler. Mystic v3 used for stylistic upscaling in Pipeline 2.
magnific.ai ↗

Claude (Anthropic)

Song creation agent — powered by a custom system prompt. Handled intake, lyrics, style descriptors, slider settings, and publish fields across all 10 tracks.
claude.ai ↗

Andor / Niamos! — composed by Nicholas Britell, Season 2 remix by Brandon Roberts. Referenced as sonic inspiration only. Wire Spine is an independent original work.

This is Part 1. Part 2 goes deeper: building a full Wire Spine album using Suno Voices — with Veyra as the lead character. How to use a video hook inside Suno itself, and what it actually takes to get a hook that holds across every generation. Stay tuned.

Wire Spine is live on Suno

Search 'Wire Spine' by Kers — all 10 tracks in sequence.
Follow the creator on X (Kers) or check out Kers on Suno.

Listen on Suno →