Modern artificial intelligence technologies are opening new horizons in architectural design. In our work on photorealistic visualizations, we use multiple tools - from hand sketches to Stable Diffusion.

We typically begin our search for inspiration in Midjourney — a neural network capable of creating photorealistic images based on text descriptions. Finding the right prompt may take considerable time, but this is the first and crucial step in forming the architectural identity. After achieving the desired result and obtaining client approval, we proceed to the next stage.

Next begins the work on creating conceptual models in Revit, one of the most popular tools for BIM (Building Information Modeling) design. This model is subsequently used at all design stages and contains not only visual but complete project information — from space planning to piping layouts.

At the architectural visualization stage, we may use Prome AI to produce high-quality renders. The neural network analyzes the Revit model and generates photorealistic images showing the project under various lighting conditions and environments. This helps us better understand and demonstrate how the building will appear in reality.

We then employ Stable Diffusion for further image refinement. This neural network enhances details, adds artistic effects, or modifies color palettes, making renders more appealing and expressive. Finally, using Photoshop, we add finishing touches by incorporating textures, adjusting lighting, and improving overall composition.

Image 1: Generated in Midjourney as a reference for further development

Images 2-3: 3D model created in Revit based on existing layouts and references

Images 4−5: Final results after processing with Prome AI, Stable Diffusion, and Photoshop

Thus, through the coordinated work of professionals and modern technologies, we achieve quality results in the shortest possible time!

Article author: Daria Grigorieva

VAU to everyone!