2024
Generative Renders
This project explores the intersection of artificial intelligence and architectural representation. A simple model screenshot serves as the foundation for an experimental workflow using ComfyUI and Stable Diffusion. Through controlled image-to-image processing and custom-trained LoRA models, the same architectural geometry transforms from photorealistic renderings to stylized architectural illustrations.
ComfyUI
Custom LoRA
Moving beyond generic outputs, the process involves training specialized LoRA models on curated datasets of art styles. This approach allows for distinct visual languages to emerge. ControlNet maintains spatial accuracy while the AI reinterprets materials, lighting, and atmosphere. The result demonstrates how machine learning can expand rather than replace architectural representation.
A Perfect Blend of Elegance and Functionality
Technical Exploration
Base geometry processed through iterative img2img refinement
Custom LoRA models trained on art precedents
Simultaneous generation of photorealistic and stylized outputs
Depth mapping and lighting control via ControlNet
Latent space interpolation for animated sequences
This workflow reveals new possibilities for architectural visualization, where AI becomes a dynamic tool for design exploration rather than just a rendering engine. The outputs showcase how neural networks can develop distinct visual languages when guided by design intent.
More Works
2024
Generative Renders
This project explores the intersection of artificial intelligence and architectural representation. A simple model screenshot serves as the foundation for an experimental workflow using ComfyUI and Stable Diffusion. Through controlled image-to-image processing and custom-trained LoRA models, the same architectural geometry transforms from photorealistic renderings to stylized architectural illustrations.
ComfyUI
Custom LoRA
Moving beyond generic outputs, the process involves training specialized LoRA models on curated datasets of art styles. This approach allows for distinct visual languages to emerge. ControlNet maintains spatial accuracy while the AI reinterprets materials, lighting, and atmosphere. The result demonstrates how machine learning can expand rather than replace architectural representation.
A Perfect Blend of Elegance and Functionality
Technical Exploration
Base geometry processed through iterative img2img refinement
Custom LoRA models trained on art precedents
Simultaneous generation of photorealistic and stylized outputs
Depth mapping and lighting control via ControlNet
Latent space interpolation for animated sequences
This workflow reveals new possibilities for architectural visualization, where AI becomes a dynamic tool for design exploration rather than just a rendering engine. The outputs showcase how neural networks can develop distinct visual languages when guided by design intent.
More Works
2024
Generative Renders
This project explores the intersection of artificial intelligence and architectural representation. A simple model screenshot serves as the foundation for an experimental workflow using ComfyUI and Stable Diffusion. Through controlled image-to-image processing and custom-trained LoRA models, the same architectural geometry transforms from photorealistic renderings to stylized architectural illustrations.
ComfyUI
Custom LoRA
Moving beyond generic outputs, the process involves training specialized LoRA models on curated datasets of art styles. This approach allows for distinct visual languages to emerge. ControlNet maintains spatial accuracy while the AI reinterprets materials, lighting, and atmosphere. The result demonstrates how machine learning can expand rather than replace architectural representation.
A Perfect Blend of Elegance and Functionality
Technical Exploration
Base geometry processed through iterative img2img refinement
Custom LoRA models trained on art precedents
Simultaneous generation of photorealistic and stylized outputs
Depth mapping and lighting control via ControlNet
Latent space interpolation for animated sequences
This workflow reveals new possibilities for architectural visualization, where AI becomes a dynamic tool for design exploration rather than just a rendering engine. The outputs showcase how neural networks can develop distinct visual languages when guided by design intent.
More Works