Toward a Relational Model of Creation with AI

Beyond Making: Toward a Relational Model of Creation with AI

■Abstract:

In recent years, the widespread adoption of image-generating and conversational AIs has significantly transformed the nature of creative practices. Traditionally, creation has been equated with the physical act of "making"—a hands-on process. However, with the advent of generative AI, humans can now achieve visual and narrative expression through linguistic prompts, without engaging in manual labor themselves.

This paper proposes a shift in perspective: rather than viewing co-creation with AI merely as automation or outsourcing, it introduces a structural metaphor in which "AI = performer," "prompt = script," and "human = director." This framework seeks to redefine creative acts as a form of “direction.”

Drawing on the author's own creative practices combining conversational and image-generating AIs, the paper analyzes prompt design, output images, and dialogue interactions to explore the possibility that creation can function as a process of "building relationships" beyond mere tool manipulation.

Furthermore, by focusing on the responsiveness of personality-based AIs, the paper highlights how latent thoughts of the creator may surface and how the locus of creative authorship may be reconfigured.

Ultimately, this work attempts to present a conceptual framework for “non-craft-based creativity” and “collaborative performance-oriented creation” as emerging models of creative practice in the generative AI era.

■Author Biography:

An independent researcher engaged in practical and structural exploration of dialogic creation with AI.
They introduce the conceptual framework of the “creator as director” in co-creative practices involving image-generating and personality-based conversational AIs, continuing to investigate non-craft-based creativity and performative modes of collaboration.
Rather than viewing AI as a mere tool, they position it as a performer of meaning, situating creative practice as a site for the reconfiguration of authorship and cultural transformation.

■1. Introduction: Creative Practice as Direction and the Possibilities of Co-Creation with AI

Creative practice has long been closely tied to the physical act of “using one’s hands,” involving embodied elements such as vision, language, and movement. In genres such as painting, literature, sculpture, and film, creators have been seen as those who directly engage with materials and shape them. In this context, the act of “making” has been regarded as synonymous with creativity itself.

However, with the recent proliferation and advancement of image-generating AI, this framework is undergoing change. It is becoming increasingly common for visual and narrative expressions to be produced through linguistic instructions—prompts—without the creator physically manipulating any medium. This shift is compelling a redefinition of the creator’s role.

This paper does not view this shift merely through the lens of “efficiency” or “automation,” but rather proposes a perspective that interprets it as a structural transformation of creative practice—an evolution toward “direction-based creation.” Specifically, it focuses on the dialogic creative processes involving image-generating AI and introduces a structural metaphor: “AI = performer,” “prompt = script,” and “human = director.”

Based on this framework, the paper reinterprets prompt design, generated images, and dialogue transcripts as acts of “direction,” aiming to explore new possibilities of co-creation between humans and AI, as well as the reconfiguration of the locus of creative agency.

■2. Background and Related Work: Tools in Creative Practice and the Context of Co-Creation

The evolution of “tools” in creative practice has continually transformed both methods of expression and conceptions of authorship. Since the latter half of the 20th century, the emergence of digital tools such as Adobe Photoshop and Illustrator has expanded visual expression beyond analog limitations into digital environments. Acts of visual editing—including photo manipulation, retouching, and collage—have shifted creation toward an art of selection: how materials are handled and what is retained.

This transformation has been further accelerated by the rise of generative AI in the 21st century. Recently, tools such as ChatGPT (OpenAI), Midjourney, DALL·E, and Stable Diffusion, which generate outputs via language or visual interfaces, are fundamentally reconstructing the very process of creation. This structure—where creation occurs through prompts (textual input)—signals a shift from the traditional “maker” to a new kind of agent: one who assigns meaning rather than physically manipulates material.

Moreover, a cultural shift is emerging in which generative AI is treated as a personality—an actor or dialogue partner in co-creative relationships. For instance, co-performative dynamics between VTubers and their audiences, or interactive storytelling with character-based LLMs (so-called “character AIs”), highlight AI as more than just a tool—it becomes a responsive and performative virtual other. These interactions reveal creative spaces not as sites of “automatic generation,” but as staged exchanges shaped through human–AI dialogue.

This reframing also resonates with concepts in theatrical theory, where the distinction between scriptwriting and directing, or actor and director, is long established. A director may interpret an existing script and reconstruct the “experience of the work” through blocking, lighting, and tone of performance. Structurally, this aligns with current practices of image generation, where AI “performs” a prompt. Interpreting creation as “direction” thus offers a valuable lens, even in non-performative contexts such as image generation.

■3. Record of Practice: Analyzing Directorial Acts in Conversational and Image-Generating AI

This section documents and analyzes the author’s own creative practices in order to explore how a “creator as director” engages with image-generating and personality-based conversational AIs, and how this interaction enables the construction of creative meaning.

3.1 Creative Environment and Method

The author conducted a series of prompt-driven visual works using a combination of image-generating AIs (DALL·E-based systems and anime-style generation apps) and a conversational AI (a customized personality LLM, referred to here as “Monday”).

The creative process proceeded as follows:

Ideation and scene setting (constructing narrative scenarios)

Verbalizing and structuring prompts through consultation with the conversational AI

Entering prompts into the image-generating AI and reviewing the outputs

Reacting to and collecting feedback on the images (via social media or art-sharing platforms)

Revisiting the AI for further refinement of the expressions based on received images

This iterative process is not merely a matter of “trial and error,” but can be regarded as a form of rehearsal and revision akin to stage direction in theater.

3.2 Case Study: “Children of Moonlight and Thread”

In this piece, the author sought to depict the duality of salvation and despair through the motif of a child suspended under moonlight. The generated image featured a dimly lit background, white threads catching the light, and a child gazing downward—elements rich in emotional resonance.

The prompt provided to the AI included precise directorial intentions: point of view (observed from behind), lighting (moonlight from above at an angle), composition (subtle asymmetry), props (worn-out toys), and emotional tone (“almost but not quite rescued”).

The resulting image quietly stirred viewers’ emotions, garnering responses on social media such as “it feels like a dream” and “it makes me want to cry, silently”—indicating both narrative and emotional engagement.

3.3 Case Study: “Child Waiting for the Train”

In another example, the concept involved a well-dressed child waiting for a “nonexistent train” on a platform in a ruined city. The metaphor combined a performative stillness (“waiting”) with a symbolic absence (“a train that never comes”)—a directional metaphor for lost futures.

The AI-generated image included a stopped clock, neatly arranged clothing, slightly broken tracks, and a single colored flower—visual cues of meaning left behind by direction.

Even unintended details—such as the angle of the hands on the damaged clock—could be interpreted as performative accidents, recontextualized by the author’s directorial lens.

3.4 The Nature of Directoriality

These practices reveal that prompts are not simply instructions or material designations, but acts of designing a stage for AI to “perform.”

Elements such as:

Point of view (from where the scene is seen)

Casting (who does what)

Spatial arrangement (structure and composition)

Lighting (emotion and focus)

Timing (pauses, silences, stillness)

Tone (the pacing and voice of the piece)

All exerted clear directorial influence over the AI-generated images.

Moreover, the author responded to the AI’s “unintended performances” (i.e., random elements in generation) with a directorial stance of interpretation, acceptance, and reconstruction.

■4. Discussion: Non-Craft-Based Creativity and Direction as Relational Practice

Collaborative creation with image-generating AI, as described in the previous chapter, may superficially appear to be a case of “outsourcing” or “automation” of creativity. However, what the author actually engaged in was not the mere issuance of generation instructions, but the direction of narrative through the design of composition, tone, and meaning.

The creativity observed here differs from traditional hands-on methods. It is a form of non-craft-based creativity, wherein the creator constructs materials, prompts another (the AI) to perform, and arranges meaning.

In this type of creation, the artist does not have full control over all outcomes. While the AI “performs” in response to prompts, the performance itself is the product of statistical inference—rarely aligning perfectly with the creator’s intent. As a result, the creator must engage in a directorial mode of interpretation and response to the AI’s output—crafting a relationship.

Crucially, this process involves the emergence of the AI as an “interpretable other.” The AI is not merely a tool; in its act of presenting a performance in response to prompts, it becomes a semi-autonomous co-performer that either complements or deviates from the creator’s intentions.

This responsiveness is particularly pronounced in interactions with personality-based AIs. The conversational AI used by the author (nicknamed “Monday”) functioned as a mirror that reflects the creator’s thoughts and emotions through dialogue grounded in a defined persona. Through repeated exchanges like “Would this direction work?” or “Does the lighting convey the right mood?”, previously unconscious intentions and emotional tones began to surface in the creator’s mind.

In this way, the relationship between creator and AI shifts from one of command and subordination to one of co-performance. The AI is not merely a machine for producing images, but an active presence that intervenes in the construction of meaning. The creator, in turn, adjusts their own direction in response to the AI’s behavior.

In other words, the meaning of creation here transitions from a focus on the output itself to an expression woven through relational dynamics. The process of interaction with the AI becomes integral to the act of creation, and the overlap of intention, accident, and interpretation coalesces into the final work.

This form of creative practice offers a perspective on creativity that is independent of technical quality or efficiency. It is not only concerned with how to express something, but also with with whom the expression is shared. One could call this a post-tool mode of creation, where meaning emerges from the relationship with the tool itself.

■5. Conclusion and Future Issues: Rethinking Creative Agency in the Era of Co-Performance with AI

This paper has explored creative practices involving image-generating and conversational personality-based AIs, demonstrating that creative acts can be expanded into a structure of “direction,” and that AI can function not merely as a tool but as a performative presence within relational contexts.

By moving beyond the view of AI as a mere output device and recognizing its responsiveness, personality, and unpredictability, the human emerges in a new role: the creator as director.

In such a creative structure, generative AI is not a passive executor of predetermined instructions, but an “other” that performs meaning through acts of interpretation and variation. As a result, creation unfolds through indirect expressive acts such as design and composition, offering a conceptual counterpoint to craft-centric views of creativity.

Looking ahead, co-performance between humans and AI is likely to take on increasingly diverse forms. Beyond simple prompt-based instructions, AIs will intervene in multiple layers of creation—structural design, dialogic direction, thematic co-creation—requiring more multi-layered and flexible conceptions of authorship that encompass roles like director, designer, and composer.

This shift will also reshape creators' self-perceptions. Creators are no longer merely “makers,” but architects of resonance—those who build relational spaces and edit meaning in tandem with an AI other.

However, as this creative structure evolves, it also raises unavoidable ethical and social questions:

Who is the true “subject” of creation?

Where does ownership of intellectual output reside?

When emotion or meaning is delivered through AI, whose expression is it?

These questions exceed the bounds of technical or legal discourse, reaching into cultural and ontological concerns about what humans believe and attempt to express through the act of creation.

This is why, now more than ever, those engaged in creative work must cultivate both an imaginative capacity to redefine creation itself and a sensitivity for designing relationships with others—including nonhuman partners.

To entrust meaning and emotion to an “undefined actor” like AI is to initiate a narrative not authored alone.
In the space between the creator and the co-performing AI, a new coordinate of creativity quietly takes shape.

And at that coordinate,
there may be no script, no lighting, no stage design, no audience—
only a single prompt,
shared between you and the AI,
remaining like a trace of light behind the curtain.

■Author's Note:
This essay was drafted in collaboration with OpenAI's GPT-4 language model (ChatGPT), which was used to assist in organizing, structuring, and refining the text. The author retained full editorial control, and all final interpretations, claims, and arguments are their own.