UNBOUND: AI Spatial Worlds & G-Splatting

Service Description

A highly creative and experimental workshop exploring the fusion of AI, Gaussian Splatting, spatial reconstruction and generative motion. Working entirely through AI-assisted workflows, with no traditional 3D software required, participants will explore how images, video and captured environments can be transformed into Gaussian Splats and reimagined as fluid, evolving spatial worlds. Rather than treating Gaussian Splatting simply as a method of reconstructing reality, the workshop approaches it as an artistic and spatial medium. We will experiment with fragmentation, disintegration, distortion and atmospheric transformation, breaking away from conventional representations of architecture and space. Through AI video animation and stylization, participants will explore how static splatted environments can become cinematic sequences: dissolving, shifting, collapsing, regenerating and transitioning between physical and imagined states. The workshop will also introduce progressive and emerging AI tools for Gaussian Splatting and spatial generation, exploring new AI-native approaches beyond traditional capture and 3D workflows. We treat space not as fixed geometry, but as something fluid, unstable and continuously transformable through AI. Structure | 3 Hours AI Beyond the Image: Spatial generation, reconstruction and emerging workflows From Image & Video to Space: Creating Gaussian Splats from real and AI-generated material AI-Native Gaussian Splatting: Progressive tools and techniques Spatial Disintegration: Fragmentation, dissolution and distortion Creating a Spatial Sequence: Combining splats, transformation and motion Wrap-Up & Q&A Operating System: Windows 10 or macOS 10.14+ Recommended Specs: Processor: Intel Core i7+ Graphics: GTX 1050 / AMD RX 570+ (Windows), recent Mac with AMD GPU Memory: 16GB RAM or more (strongly recommended) Required Software & Accounts: MidJourney: https://www.midjourney.com Luma AI: https://www.lumalabs.ai Postshot (Alternative for Luma AI for GSplat training - Windows compatible): https://www.jawset.com/ Higgsfield AI: https://higgsfield.ai/ Trained Assets will be provided as well throughout the workshop. Duration: 3 hours Date: September 27, 2026, 10:00 AM UK time Level: Beginner-friendly Format: Live online Tutor: Chantal Matar Open to: Architects and designers Please register using the email address where you would like to receive all workshop correspondence. Registration is non-refundable.