AI Playground
Welcome to my Playground
This is where I experiment with the cutting edge of technology—tools, platforms, and ideas that push creative and professional boundaries. From AI image generation, editing, 3D, audio to immersive AR/VR experiences with Apple Vision Pro, real-time rendering, and interactive prototypes for apps and interfaces—this space is all about exploration.
Some of these experiments evolved into professional solutions for clients, while others remain pure passion projects. The AI Playground is a window into that process: where imagination meets innovation, and tomorrow’s workflows take shape today.
Real estate pitch
A fast-turnaround pitch concept built in just a few hours, starting from a single watercolor illustration of a house.
Workflow
I transformed the illustration into a photorealistic image using Flux Kontext and crafted a detailed close-up of the brick wall with Nano Banana. Combining these assets into start and end frames, I generated a dynamic zoom-out animation with Adobe Firefly that captures convincing depth—something that would typically take days to achieve in 3D. I’ve also animated the bricks collapsing in Cinema 4D with MoGraph, leveraging traditional 3D tools where AI still falls short.
the impossible roto
Designed and animated this video within just two weeks, and the mandatory stroboscopic effect with multiple frames on screen didn’t made thing easier.
Workflow
I tested several established VFX roto workflows, including the latest AI-assisted segmentation tools, but the subject’s rapid movement meant frame-by-frame masking (a prohibitively time-consuming thing) would be my only choice. The breakthrough came while using AI depth maps to blur backgrounds on another project—I realized I could use the depth pass to isolate the subject from the environment. This approach worked seamlessly, requiring only a handful of manual masks to separate her from the floor, dramatically accelerating the process.
Check the final project here.
augmented reality
We were commissioned to adapt a VR app built in Unity 3D for Meta Quest, that looked like an retro game, into an elegant experience with native Apple Vision Pro augmented reality application. The project required overcoming a steep learning curve, compounded by Apple’s limited documentation and the absence of established forums or tutorials. Additionally, we needed to build a brand-new 3D map environment from scratch, since using real-world elevation data (e.g., from Google Earth) posed potential legal risks for the client.
Workflow
Procedural Terrain: I generated the base terrain with GeoGen, a procedural tool I learned on the fly. The default textures lacked quality, so I rendered multiple perspectives of the terrain, enhanced them with AI upscaling for fine detail, and combined them through projection mapping to produce high-resolution, seamless textures.
Map Locations: For visitable VR locations, I rendered fisheye views of the landscape, then refined them with AI to achieve a polished, immersive look.
Splash Screen: Starting from a Midjourney concept of a house in the mountains, I integrated it into the custom 3D landscape and enhanced it with AI to match the lakefront setting of the map.
3D House: Using Tripo, I quickly generated a proportionally accurate base mesh of the house. This AI-assisted model allowed me to efficiently re-model and optimize a lightweight version in Cinema 4D.
Vibe coding
When I realize a missing tool could make my work easier, I turn to Cursor to prototype solutions and see how far I can take them.
The client needed a straightforward way to review a virtual environment remotely. To solve this, I built a lightweight HTML viewer that allowed them to open a fisheye 360° render in any browser—without distortion and without the need for additional software.
Here is a link to the file.
While working on the AIGA GenUX Symposium project I had this idea of building a browser-based ASCII video generator, that converts video into fully customizable ASCII animations in real time.
Here is a link to the file.
