Mallory Vice

In the artistic world, artificial intelligence is often discussed either as a threat to creativity or simply as a new technological tool. Yet a different possibility may emerge: a new relationship between human imagination and machine-scale generative systems in which creativity evolves through continuous interaction between human artistic judgement and artificial intelligence.

Rather than viewing AI as an autonomous creator or as a mere software tool, the process functions as a reciprocal system in which both human and machine contribute different forms of capability to the evolution of a work.

Within this framework, the AI does not simply produce finished compositions. Instead, it operates as a large-scale variation engine capable of generating enormous numbers of related musical possibilities. Through iterative prompting, stem generation and transformation, the system continuously expands the available creative landscape. The human creator then acts as curator, selector, arranger and aesthetic guide — identifying promising structures, refining emotional direction, preserving coherence and shaping the long-term artistic intent of the work.

Variations are generated, successful characteristics are selected and useful traits are carried forward into subsequent iterations. The process is directed through conscious human judgement. Harmonic language, narrative pacing, instrumentation, texture and emotional impact are continuously evaluated and reshaped according to artistic intention.

This creates a form of guided generative evolution in which the final work emerges not from a single act of composition, but from an ongoing dialogue between human intentionality and machine-assisted possibility generation. 

As systems improve, this methodology may extend far beyond music into film, visual art, literature and immersive media. Some artists may increasingly choose to operate less as direct constructors of every detail and more as architects of evolving generative environments — cultivating, selecting and directing creative emergence across vast possibility spaces that no individual human could manually explore alone.

The artist of the future may be distinguished not by the ability to manually create every component of a work, but by the ability to guide the evolutionary development of creative systems toward meaningful artistic outcomes.

Through iterative prompt evolution, stem extraction, recombination and editorial refinement, the human creator continuously guides the developmental pathway of the music. Human imagination and machine-scale generative capability operate together.

Whereas Mallory Vice remains rooted in traditional song-based rock composition with AI-supported production, Pansarion is an experimental platform vehicle for Directed Generative Stem Composition.

Stem Composition workflow

The process begins with the creation of a primary musical prompt with human lyrical input within an AI music generation platform such as Suno or Udio. This primary prompt establishes the stylistic and musical identity of the work, including genre, instrumentation, harmonic language, tempo, rhythmic feel and overall production aesthetic.

Secondary and tertiary prompt variations are then created from the original source prompt. These derivative prompts maintain shared musical characteristics — such as key centres, chord progressions, tempos, instrumentation palettes and production styles — while generating multiple alternative musical performances, arrangements and melodic interpretations. The result is a large family of musically related source tracks rather than a single isolated composition.

Selected versions are then separated into individual stems. These stems may include drums, bass, guitars, keyboards, orchestral layers, lead vocals, backing vocals and atmospheric elements. In large projects this can result in between 100 and 200 individual audio stems.

The stems are imported into Logic Pro, or an alternative Digital Audio Workstation (DAW), where a first-stage compositional edit takes place. During this stage, stems are:

Arranged into larger musical structures,

Trimmed and synchronised,

Duplicated and layered,

Combined across different AI generations,

Dynamically edited to construct a coherent suite or song form.

At this point the project transitions from AI generation into detailed human-directed composition and production.

Vocal stems are then extracted and processed in a vocal platform such as Audimee, Kits AI and ACE studio, where alternative vocal models can be applied. Melody, phrasing and harmonic structure may also be modified using internal editing tools to refine performance character and vocal arrangement.

Instrumental stems are selectively transferred into an instrument platform such as ACE Studio, Synthesizer V Studio and Vocaloid 6, where instruments can be replaced, transformed or re-performed using alternative synthesis or performance models. Additional editing of melody, harmony and articulation is also be carried out with internal editing tools.

The newly generated stems are then re-imported into Logic Pro for a second-stage production process. This stage includes:

Detailed arrangement refinement,

Harmonic balancing,

Layering,

Automation,

Effects processing,

Spatial positioning,

Transition editing,

Structural revision,

Dynamic shaping.

Finally, all stem tracks are mixed into a stereo master within Logic Pro. Mastering processing is then applied to produce the finished musical work, including equalisation, compression, stereo enhancement, loudness optimisation and final tonal balancing.

The resulting composition is therefore not a direct AI-generated song, but a  human-directed production constructed through iterative generative sourcing, stem-level recomposition and multi-platform editorial transformation.

Directed Generative Stem Composition (DGSC)

Directed → the process is controlled by intentional musical decisions rather than random generation.

Generative → AI systems are used to create source material.

Stem → the workflow is based around extraction, manipulation and recombination of stems.

Composition → the end result is an authored musical work rather than an AI output dump.