Elemental Alchemist: A Generative Interface for Semantic Control of Particle Systems Across Dynamic Levels of Abstraction Conference: Preprint; ; CCS: Human-centered computing Interactive systems and tools CCS: Computing methodologies Animation

Kyzyl Monteiro, Evan Atherton, George Fitzmaurice, Qian Zhou · 2026

Read Elemental Alchemist: A Generative Interface for Semantic Control of Particle Systems Across Dynamic Levels of Abstraction Conference: Preprint; ; CCS: Human-centered computing Interactive systems and tools CCS: Computing methodologies Animation by Kyzyl Monteiro, Evan Atherton, George Fitzmaurice, Qian Zhou (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-10014.txt.

Introduction

Monteiro and three colleagues built Elemental Alchemist, a generative tool for meaning-based control of particle systems across levels of detail. Editing particle visual effects matters for digital stories, but results artists can steer remain hard to reach. Users must find the settings that match their own goals. That task needs a clear grasp of the system and how settings map to high-level intents, such as making a fire look angry.

The system has two parts. A scene-based brush palette makes tools based on scene context. A generative panel shows key low-level settings and turns them into mid-level meaning traits and high-level idea-level controls. A test with 10 novice and 5 expert VFX artists showed the system helped users reach their intents.

Catalog record

CREATORKyzyl Monteiro, Evan Atherton, George Fitzmaurice, Qian Zhou
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE113 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.10014

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Elemental Alchemist: A Generative Interface for Semantic Control of Particle Systems Across Dynamic Levels of Abstraction Conference: Preprint; ; CCS: Human-centered computing Interactive systems and tools CCS: Computing methodologies Animation Kyzyl Monteiro, Evan Atherton, George Fitzmaurice, Qian Zhou Abstract . Editing particle-system visual effects (VFX) is vital for digital storytelling, but achieving controllable, art-directable results remains challenging due to their multi-dimensional nature. Given a large collection of parameters, users must find the ones relevant to their creative goals—a task that requires a systematic understanding of the particle system and how parameters map to high-level intents, such as making a fire look angry. Elemental Alchemist is a generative interface that transforms user intent into contextualized controls for semantic editing of particle systems. The system introduces two components: a contextual brush palette that generates tools based on scene context, and a generative control panel that surfaces relevant technical parameters and abstracts them to generate mid-level semantic attributes and high-level conceptual controls. An evaluation with 10 novice and 5 expert VFX practitioners shows the system supported users in translating high-level creative goals into particle system parameters. Figure 1 . Elemental Alchemist proposes a generative interface for semantic control grounded in task context and user intent. It generates a contextual brush palette with scene-aware tools for sketch-based semantic edits. From a sketch or text prompt, it generates a control panel that decomposes intent into synchronized controls across three abstraction levels: conceptual, semantic, and technical. Together, these components help translate high-level creative goals into low-level parameter edits. 1. Introduction Particle-based visual effects (VFX) are fundamental to digital storytelling. Spanning natural phenomena such as fire and water, as well as fantastical elements like magic, they contribute realism, atmosphere, and spectacle. Even when used decoratively, particle effects are rarely incidental; they are guided by the vision of storytellers like creative directors and VFX artists, and play an essential role in shaping the ambience and narrative meaning. Yet, designing, editing, and controlling particle effects remains challenging, specifically technically demanding. Designers think in semantic goals—e.g., “make the fire look angry” or “direct the fire towards the tent”—but must work through low-level parameters (emission rate, velocity theta, particle lifetime, etc.). Bridging this semantic–parametric gap is difficult: parameters have opaque names, contribute in indirect ways, and often force reliance on trial-and-error, leading to limited controllability as designers remain uncertain which controls, values, or ranges will achieve their goals. As a result, novices face a steep learning curve and struggle with art-directability —translating creative visions into concrete parameter adjustments. Addressing these challenges requires VFX tools to be scene-specific, allowing designers to work directly on what they see rather than navigating generic options. Additionally, controls should enable smooth transition between broad, high-level concepts (like “make it playful”) and fine-grained details (like adjusting a specific velocity value), while keeping those layers of control consistent with one another. Guided by these design requirements, we present Elemental Alchemist, a generative interface for semantic control of particle effects. The system introduces two components: (1) a contextual brush palette that generates scene-aware brush tools, allowing users to express scene-based intentions, for example, pushing fireworks with a gust of wind or reducing a burst near a rooftop. These brushes are generated dynamically based on the scene, ensuring that the available tools remain relevant and interpretable; and (2) a generated control panel that translates a user text prompt into a hierarchy of controls spanning three levels of abstraction: conceptual, semantic, and technical. For example, given a natural language goal “make it more playful” , the system generates high-level conceptual controls— vibrancy and movement style , which are then decomposed into mid-level attributes such as opacity variation , burst timing , and movement variability . These mid-level attributes are linked to low-level technical parameters like emission rate and velocity theta . Changes at the high level update mid and low level parameters automatically, while adjustments at the mid or low levels are synchronously reflected upward to maintain coherence. Together, these components allow users to create effects using descriptive, meaning-based terms that match how they are imagined, while also giving them the flexibility to intervene at any level of abstraction. At the same time, all controls are grounded in precise, low-level parameters of a deterministic, parameterized particle system engine. This work makes three contributions: (1) The design and development of Elemental Alchemist , a generative interface for semantic control of particle systems, a highly parametric creative domain. (2) A mechanism for generating controls at multiple levels of abstraction, with synchronization of controls across the abstraction levels. (3) Implications and empirical findings from two user studies with both novice users (n=10) and experts (n=5) that demonstrate how generative interfaces support creative control over particle effects. Our findings show that the system supported novices in rapidly translating high-level creative goals into particle system parameters, while also scaffolding their spatial reasoning and enabling intuitive exploration. Novices adopted three distinct prompting strategies—outcome-oriented, workflow-oriented, and tool-oriented—but found mid-level abstract controls to be the most effective entry point, fluidly moving up or down to high- and low-level controls as needed. Experts appreciated the system’s ability to translate intentions, filter parameters, and generate hierarchical controls that aligned with their existing workflows, while highlighting its potential to bridge communication gaps between technical and creative stakeholders. 2. Related Work To situate this work, we draw upon three areas: particle effect design tools, which establish interaction paradigms for this domain; generative user interfaces, which demonstrate that interfaces can be generated from context; and ladders of abstraction, which provide a conceptual foundation for multi-level control. 2.1. Particle System Designing Tools Particle effects have long been central to digital storytelling and visual expression, since early work on particle systems ( 50 ) . Commercial authoring environments such as Houdini, Maya Bifrost, Unity, and Unreal Engine have provided node-based interfaces and solver primitives for professional-grade effect creation ( 3 ; 56 ; 63 ; 22 ) . While powerful, these tools remain parameter-heavy and demand deep technical expertise. Prior work has explored alternative interaction paradigms to reduce barriers to expressivity and control in digital creation. Draco and Kitty introduced kinetic textures, enabling users to animate illustrations by sketching motion fields that define movement patterns ( 36 ; 35 ) . Energy Brushes extended the painting metaphor to elemental dynamics such as fire, smoke, and splashes, allowing artists to specify underlying forces through broad strokes that drive particle behaviors ( 71 ) . RealityCanvas brought this painting metaphor into spatial contexts, supporting embodied sketching of animations in physical space and enabling improvisational exploration in situ ( 68 ) . MagicalHands further demonstrated the potential of mid-air gestures for particle control, focusing on embodied interaction within immersive environments ( 2 ) . Building on these metaphors of brush tools and energy-based design for particle effects, our work advances this direction by generating brush tools dynamically based on scene context such as the particles, objects, and energies currently present in the scene. Most recently, KinemaFX reframed the challenge of designing particle effects as a search problem, combining semantic descriptions and kinematic abstractions to enable users to explore large libraries of pre-made effects ( 73 ) . After finding suitable effects, users compose them using fixed transformation controls—position, scale,

Read the complete text