This sprint took our first steps into AI-generated immersive content, delivered through two working executions. The objective was to prove out AI world-building as a pipeline accelerator, reducing world-building time for Studio teams while opening up new content formats in The Lighthouse. Using Marble by WorldLabs, we generated synthetic Gaussian Splats from AI images, brought them into Unreal Engine via a Marble plugin, and rendered the resulting 3D environments into the Originals Theatre through Disguise. Alongside this, we built a simple pipeline to take an AI-prompted 360 image or video and map it directly into The Lighthouse. Both demos were documented as repeatable workflows, built by artists, for artists, and designed as a springboard. The result demonstrated that AI-generated 3D environments can meaningfully accelerate production, with Gaussian Splats and AI generation proving viable as pipeline tools at the scale of our immersive spaces.

This experiment set out to explore how AI could be used to rapidly productise a shared, browser-based interactive experience on a non-standard physical canvas; in this case, the display setup within our own bar. First we translated the physical space into a set of rules an AI could work with. We built a calibration system to map the real-world parameters of the canvas (the screen resolution, shelf positions, and LCD placement) and fed these to AI as constraints. From there, we prompted it to generate a game that operated within that specific framework: gravity behaviours, collision logic, and playable boundaries defined by the physical environment itself. Using Qwen Code, we rapidly scaffolded a websocket server and a database for scores and leaderboards. The result was a retro-style side-scrolling coin collector game, with 60-second rounds, and a live leaderboard, playable by scanning a QR code on their phone and playing in browser. The exercise demonstrated how quickly a bespoke, canvas-aware interactive can be created once the framework is established, and how AI can abstract the heavy lifting of game logic and asset creation when given clear parameters.

A client requirement prompted us to investigate the viability of Gaussian splats as a practical, low-budget alternative to traditional spatial scanning. We tested three capture methods side by side: a budget iPhone LiDAR scanner, a raw iPhone camera feed, and an 8K 360° camera, processing each through tools including Splatica and VPS. The strongest output came from the 8K camera processed through Splatica. We applied this method to a 10-minute video captured inside The Lighthouse, using it as a stress test for low-light performance, processing speed, and in-browser navigability. The result was a fully walkable 3D space delivered as a browser link. We further experimented with adding camera points to the scene to produce animated walkthroughs and flythroughs. From a 10-minute capture to a processed output in just two hours, the exercise demonstrated how quickly we can create a 3D representation of a space; opening up applications in client presentations, site survey reports, and live construction updates.

The genesis of this project was a desire to acquire foundational knowledge of Unreal, encompassing lighting, editing, integration of the Niagara system, and Megascans. My inspiration was to create an environment reminiscent of the aesthetics of “The Last of Us” video game, characterized by beauty amid the aftermath of an apocalypse. The objective was to test my capabilities in a limited timeframe: Once I had assimilated the rudimentary principles of Unreal, I produced a video within a day, primarily employing Unreal as the key tool, complemented by a 10% utilization of AE to enhance its visual impact.

Initially branching off from our original ‘Room to Breathe’ in Unreal, this experiment encompasses lighting, integration of volumetric systems and Megascans assets. The objective was to develop a secluded atmosphere in the skies, utilizing Ultra Dynamic Skies systems for inspiration. My aim was to repurpose elements from the original project to create a completely unique environment. The completed video was rendered and graded within Unreal Engine, requiring no external resources, and was finished in just half a day.

Magda is another of the portraits created from a month-long period of experimenting with Nightcafe, an AI generator that I found incredibly intuitive and addictive. For Madga I used a template provided by Nightcafe that uses the style of Greg Rutkowski, Artgerm and WLOP as references. I appreciate the intimacy of the resulting portraits, which differ from the more conceptual art style often seen in generated images. Nightcafe proved to be an excellent tool for concept design, given the incredible quality and quantity of work one can produce in a short amount of time. It’s truly remarkable.

This clip was filmed pre-build in our warehouse, while we tested the infrared tracking camera system. We were testing the overall system latency, and exploring the tile interactive mechanic. The tiles flip when the camera sees motion: tiles that haven’t been ‘affected’ by motion play video one; and tiles that have been ‘affected’ play video two, which is a much higher contrast. During this test, Rupert, the artist, was able to finalise the video layers and we then went into Beta testing on-site.

After working on a project for a the high-end jewellery brand, I had diamonds on my mind, so I began exploring the inside-a-diamond environment. To achieve this effect, I played with prismatic reflections and subsurface scattering in Cinema 4D. This software, along with After Effects, has been instrumental in creating the desired visual outcome. Through experimenting with these techniques, I have been able to produce striking visuals that capture the essence of the diamond’s beauty from within.

I started by using the Niagara Particle system in Unreal to showcase my learnings from Maxim Zhestkov’s work. I used Megascans assets to create a similar look and feel and created the environment within Unreal, using lighting, volumetric systems, and editing tools to achieve a cohesive look. Everything was rendered and graded in Unreal without external software. I gained valuable experience with particle systems and environmental design.

I had been working extensively with ambient soft glowing particles, but I wanted to experiment with something that had a little more personality and movement. This led me to create a unique light that appears to be evading from a swarm of fireflies, adding depth and intrigue when viewed on a phone screen. As someone who had been working in a real-time environment for quite some time, I was eager to explore the possibilities of lighting and materiality within Cinema 4D. To bring this project to life, I utilized X-Particles to add additional life and secondary lighting with captivating motion. Overall, the project was executed using several software programs, including Cinema 4D, Redshift, and X-Particles.

Svenja is the second portrait I created from a month-long period of experimenting with Nightcafe, an AI generator that I found incredibly intuitive and addictive. I appreciate the intimacy of the resulting portraits, which differ from the more conceptual art style often seen in generated images. Nightcafe proved to be an excellent tool for concept design, given the incredible quality and quantity of work one can produce in a short amount of time. It’s truly remarkable.

We carried out this sprint to explore how Leap and Looking Glass can be interfaced together for the use of a luxury watch or jewellery brand. We wanted to see if we could use Leap to create a user interface on Looking Glass, allowing brands to configure a watch so that customers can try it on virtually. The interface worked well, but would need efficient call to actions to make the experience super intuitive.

I began by delving into the latest 5.1 updates in Unreal, specifically Lumen reflection, and incorporating megascan assets. The objective was to establish a platform to exhibit the scene’s lighting and reflections solely utilizing Lumen, while rendering directly to apple pro res. The entirety of the project was rendered and graded within Unreal Engine, demonstrating the remarkable outcomes that can be achieved.

The aim of this experiment was to synchronize sound and image to generate a captivating audio-visual experience. Using Cinema 4D, a visual effects software, I employed music and beats through mograph to procedurally animate the graphics, resulting in an immersive visual soundscape.

This clip was filmed on-site at, Outernet London. It shows our temp delivery of the experience’s interactive element. We were testing the scaling of the portal, and the playback of the pre-rendered content behind the real-time layer. During this test, we decided on the size of the portal, the number of tiles and the distance between each one. We wanted to ensure that two audience members could interact simultaneously.

I embarked on this project to experiment with more real time tools in various spatial ad lighting set-ups to see how they would react. I played around with Volumetric Animations, First using Ember Gen to generate realistic smoke and fire simulations and then using Unreal Engine to build a 3D World for it to sit in with reactive volumetric lighting.

Over an extended duration, I’ve dedicated my time to thoroughly exploring Midjourney. The introduction of Midjourney’s highly anticipated ‘zoom out’ feature prompted my curiosity, leading me to explore its potential over a couple of days. After getting the Midjourney images I upscaled them, using Topaz, and then built the zoom animation in After Effects. It’s a cool effect that possesses the potential for a meaningful application, particularly when integrated with a well-suited concept.

In The Lighthouse we have our Generative AI experience that can transform you into whatever you wish right before your eyes in real-time. The experience uses an integrated camera, Comfy UI, a real-time optimised stable diffusion model, and the text prompts entered on the touch screen. As the person moves in front of the camera, the model rapidly generates new images that react to the changing pose, position, and lighting. With exciting additions in the works, watch this space to see how it develops.

Which VR headset reigns supreme? We have both headsets in our London office, and we’ve been taking some time to explore the immersive experience potential of the two, focusing on content creation, interactivity, pre-visualisation and workflow. But which comes out on top? head to the Writer's Corner to find out!