Spatial Infill & Adaptive Reuse



Overview: This research demonstrates an "Adaptive Infill" workflow—leveraging Lineart ControlNet to detect existing spatial boundaries from a raw photograph of an empty interior and simulating various high-fidelity programmatic outcomes.





01. Nordic Minimalist Simulation

  • The Control: ControlNet: Lineart Coarse (Weight 0.45). By isolating the structural edges and window apertures of the empty room, I established a rigid geometric container that prevents spatial distortion during the generation process.

  • The Prompt: Modern apartment, Nordic style, cinematic film still. The focus was on environmental mood-boarding, exploring how soft natural light interacts with minimalist Scandinavian aesthetics.

  • The Result: Environmental Mood-boarding. Using DPM++ 2M with a Karras schedule, I produced a high-detail visualization that respects the original site's architecture while injecting a professional "high-budget" cinematic atmosphere.

02. Retro-Modern Materiality & Masterpiece Furniture

  • The Control: ControlNet: Lineart Coarse. Maintaining the same structural weight (0.45) ensures that the building's skeleton remains constant while allowing the AI to "fill" the space with diverse furniture typologies.

  • The Prompt: Nordic Retro Style, Masterpiece Furniture, shallow depth of field. This experiment moves beyond basic styling into curated interior storytelling, emphasizing high-end materiality and furniture curation.

  • The Result: High-Fidelity Programmatic Testing. The output demonstrates a sophisticated blend of vintage textures and modern lighting, showcasing the ability to rapidly test different brand identities within a single physical site.