Two Stages of a Conversation With a Fictional Machine: Gallery Installations and Consumer Character Platforms

The Gallery Stage: A Character Occupies a Shared Room

A gallery conversation with a fictional machine begins with a spatial decision. The character might appear as a projected face on a wall, an animated figure distributed across several surfaces or an abstract visual presence responding to visitors. The audience does not necessarily approach the work one person at a time. People enter, leave, observe each other and interrupt the interaction simply by occupying different parts of the room.

That shared environment changes the technical requirements. A camera-reactive installation must account for the location of the sensing area, the visibility of projected content and the number of simultaneous visitors. A character that responds to every person independently can become visually incoherent when several visitors enter the interaction zone. A better design may assign the room one collective state and use clear visual transitions to communicate changes in attention.

The Consumer Stage: A Character Appears Through a Personal Interface

Consumer character platforms present a different arrangement. The fictional character appears within an interface controlled by an individual, and interaction may move between text, audio and visual media. The room itself is no longer the primary sensing and display environment. The application organizes the encounter, controls the available interaction methods and presents the character through a personal device.

TheFriend is one consumer-side example: an adults-only platform featuring fictional AI characters, text chat and vertical short films of approximately one minute. Its character catalog illustrates how a fictional personality can be presented through a collection of interface-based media rather than a physical installation. This provides a useful point of comparison for artists considering how a character's identity is distributed across different forms of interaction.

Define the Character Before Building the Interaction

A convincing installation does not begin with a camera or a language model. It begins with a character specification. Define what the machine can perceive, what kinds of responses it can produce, how it acknowledges visitors and what happens when it cannot interpret an interaction. A character that appears to recognize people should communicate the limits of that recognition. A projected figure that reacts to movement should not imply that it understands a visitor's intentions when the system only detects position or motion.

Separate the character's artistic identity from its technical capabilities. A character may have a name, visual language, voice and conversational style, while its sensing layer provides only a limited set of input variables. This distinction is important when using generative models. Model-generated dialogue can vary, but the installation still needs stable rules for duration, inappropriate input handling, output length and recovery when an external service becomes unavailable.

• Write a short character description and interaction boundary.
• Specify which visitor actions the system can actually detect.
• Define the response format for each interaction state.
• Establish a consistent fallback personality and visual state.

Connect Sensing, Generative Logic and Projection Mapping

A practical interactive installation tutorial should separate sensing, interpretation, content generation and display into distinct stages. For a camera-reactive installation, the camera captures an image, a vision process extracts a useful signal, application logic decides what the signal means and the renderer creates the response. Keeping these stages separate makes it easier to replace a tracking method without rebuilding the entire artwork.

Projection mapping adds a spatial transformation between the rendered content and the physical surface. Establish the target geometry, projector position, focus and environmental lighting before finalizing the interaction design. For example, a character might move between three architectural surfaces in response to a visitor approaching different zones. The interaction logic should identify those zones consistently, while the mapping system maintains alignment independently of the character's conversational behavior.

• Define the sensing area and expected visitor positions.
• Convert sensor data into stable interaction events.
• Keep character logic separate from display geometry.
• Reserve a fixed rendering state for calibration and maintenance.

Design Conversation for Shared Attention and Limited Time

A museum interactive exhibit operates in a social environment. Visitors may spend only a short period with a work, arrive in groups or watch another person before deciding whether to participate. A long conversational sequence can exclude people who arrive halfway through it. Use recognizable interaction states and short response cycles so that newcomers can understand the work without waiting for an entire conversation to finish.

Give the audience a clear indication of whose action currently influences the character. This could be a highlighted floor zone, a change in projected composition or a visible transition between individual and collective attention. If multiple people are present, establish whether the character responds to the nearest visitor, the first detected visitor or a combined group signal. Ambiguous priority rules create confusing behavior even when the sensing hardware works correctly.

Maintain the Difference Between a Fictional Character and a Responsive System

The character's presentation should make the fictional premise legible without overstating the system's capabilities. Visitors should be able to distinguish an authored response, a generated response and an automated reaction to movement. This distinction becomes especially important when a character appears to remember previous interactions or addresses individuals directly. Explain the role of sensing and generation in language appropriate to the exhibition, and establish clear boundaries for what information the installation retains.

For touring work, document the character specification alongside the technical configuration. Record the approved visual assets, response constraints, sensor positions, projection surfaces, content dependencies and fallback behavior. A venue that changes the room geometry or replaces a projector may alter how the character appears, even if the application remains unchanged. The artistic team should identify which elements are essential to the character and which can adapt to a new architectural environment.

Choose the Right Stage for the Intended Encounter

A gallery installation is suited to encounters in which space, collective presence, physical movement and architectural scale contribute to the meaning of the work. A personal character interface is suited to an interaction organized around an individual's screen and chosen media. These are different production contexts, not interchangeable delivery formats. Moving a character from one context to another requires decisions about pacing, privacy, input methods and the relationship between audience members.

The production brief should state what the audience is expected to do, how long an interaction should take and what remains visible when nobody participates. It should also specify whether the system needs network access, which equipment must travel and what the venue must provide. Once those requirements are established, create a small interaction diagram and a room layout showing the character's visual presence, sensing boundaries and visitor circulation. Use those two documents as the starting point for the production specification.

Questions readers ask