large worlds with complex ai

every world, more alive

Science Fiction Assets & Cutting Edge AI: (1.) megastructure environment design, (2.) cyborg character/creature design and animationm, (3.) ai/ml for generalist agents (LLM to speech/action) (4.) ai/ml for terraforming swarms (procgen/drl + agents)

Our agency is focused on asset design on the one hand and ai design on the other hand. We posit that the newest and most advanced science fiction worlds also require the newest and most advanced AI implementations. The assets we design are (1.) science fiction megastructures and environments and (2.) cyborg humanoid and non humanoid characters (and animations). The ai we design are for (3.) generalist agents that can extemporize in dialogue and action and (4.) swarms of agents that can be used to transform worlds in real time (build buildings, modify terrain, etc.). We are often brought in on game studio and game tech projects to contribute design pieces or to expand/implement new AI systems. We are known for our meticulous attention to detail, unique style and embrace of experimental techniques. Every world we touch feels more alive than it was before.

assets

ARCHITECTURe, Levels: megastructures

The studio offers architectural modelling specifically tailored for virtual spaces (web, game worlds). When custom models are needed that express the ideas, concepts, themes and sensibility of a specific world, we design that for you. Our studio is most adept at large scale stylized science fiction installations (space stations, megastructures, futuristic cities, etc.).

Characters, Animations: cyborgs

The studio has extensive experience in creating, rigging and animating science fiction characters and creatures, humanoid and non-humanoid. By request, these creatures can also be configured to be used with motion capture (for cut scenes), or live motion capture (for performance).

ai

generalist agents: llms for dialogue and actions

The studio has great familiarity and depth with the InWorld API, which allows one to use Large Language Models (LLMs) to drive NPC behavior in games. The studio is also familiar with custom deploying multi-modal LLMs for character sensing and behavior, allowing for more complex and open ended strategic planning.

robotic swarms: game AI+ml for real time terraforming

The studio has developed various methods for real-time architectural generation in game by agents; these agents generate chunks of architecture in real time (dynamic, evolving architecture). Additionally, the studio has created agents that program themselves/generate new behaviors with reinforcement, imitation learning, etc.

work history

Current & Previous Clients

Agile Lens, Australian National University, Goethe Institute, New Art City, Anti-Eviction Mapping Project, Wild, Gray Area, Buzzfeed, Mary Mattingly, Richard Rendering

select CASE STUDIES

[VR, Level Design] Christmas Carol VR for Agile Lens, Level Studies, 2023

We created a stage set for Agile Len’s VR production of Christmas Carol by Charles Dickens. We created a Victorian London “level,” a variety of “sets” (like the Scrooge bedroom above) and also experimented with a variety of complex volumetric fog transitions.

[VR, Data Visualization] Climate Data Visualization VR for Australian National University, 2024

We created a VR data visualization for the Australian National University. The program visualizes “guassian splats” and allows the user-scientist to manipulate, scale, transform, rotate and explore the scan. These scans are being used to train a machine learning model to predict how much carbon sequestration different types of terrain achieve.

[ML, Dialogue] LLM Dialogue via Inworld for Richard Rendering, 2024

In this demo world, you can see a short example of talking to one of the characters, Akira. We didn’t design any of the aesthetics or characters. We implemented the InWorld dialogue system. We also made it so that the character can go to waypoints, and that worked well with the dialogue system. We are currently exploring an integration with vision-gpt for better contextual awareness.

[AI, Behavior, Robotics] Robotic Simulator for Grizzly, 2024

We created a highly complex simulation of an excavator. The excavator has ISO controls, and is able to simulate digging in the ground (via the voxel plugin). The excavator is also able to save both articulation and camera data that was used to train an imitation learning model to dig a trench autonomously.

[Music Video, Level Design, Reinforcement Learning, AI Texturing] Music Video for Artist RTyler, 2023

We created a music video for RTyler, using a combination of reinforcement agents and a reskinning with parts of our music video with stable diffusion. The video is meant to represent an advanced civilization that has radically fused biology and computation, and is able to generate all manners of creatures and agents.

[Level Design] Megastructure Studies for New Art City, 2020

We created a series of “megastructure studies” for New Art City, helping them to determine more about what kinds of scales of architectural space are helpful for their platform. The studies were also an advanced survey of importing grasshopper generated forms into the platform.

Tutoring, teaching & training

The agency also offers small group tutoring, subject specific workshops and trainings for teams and individuals in the above subjects.