How It Works

Using Aperture Narrator is different from chatting with an AI. Instead of issuing prompts and accepting whatever emerges, experiences are shaped through care and constraint — much like cultivating a bonsai tree.

BONSAI is the name we use for this curatorial practice and the tooling that will support it.

The Pot

Creation begins with human-defined canonical information. This foundation establishes the boundaries within which generative systems operate.

These boundaries extend beyond simple facts. They include locations, themes, events, tone, and institutional guidelines that shape what can be created. By defining the container first, experiences can vary without losing voice, accuracy, or intent.

Close-up of two hands holding soil with a small green seedling growing out of it.

The Seed

Once the boundaries are set, the experience itself is shaped.

Story arcs, main and supporting characters, modes of interaction such as puzzles or exploration, and other narrative elements are defined at this stage. These choices give direction to the system, allowing generation to align closely with the creator’s goals rather than emerging arbitrarily.

A bonsai tree in a shallow pot, with a twisted trunk and lush green foliage, illuminated by warm sunlight.

Within this framework, the engine generates experiences with controlled variation.

Multiple paths, outcomes, and presentations can be produced to suit different audiences or contexts, while remaining anchored to the same underlying canon. This process is not an open-ended conversation, but a cultivated form of generation designed to minimize drift and avoid the introduction of unverified or unintended material.

Growing

Hands trimming a bonsai tree with scissors in a pot.

Pruning

At every stage, human authority remains intact.

Creators can revise, regenerate, refine, or remove any portion of the output. Nothing is final until it is validated. In this way, authorship and accountability stay with the institution, rather than being deferred to the training data or tendencies of a model.

Sharing

When an experience is ready, it can be shared in a variety of ways depending on the desired context and audience.

Previewing is always possible, ensuring that what is published is exactly what was intended. The result is an experience that feels alive and adaptive, while remaining grounded in care, oversight, and trust.