Note: This article was written and published in a different language and is a translated version of the original published here.
01 — Generating the space
The project began with a prompt to define a contemporary Mediterranean dwelling organized around a central courtyard.
Rather than looking for a final image from the beginning, the intention was to obtain an explorable 3D environment that could become the setting for a small architectural narrative.
The space generated thus functioned as a starting point: a base on which to begin to make decisions about composition, route and camera.
02 — Camera direction in 3D Editor
Once the space was generated, I used the 3D Editor to walk through the architecture and start laying out the different shots.
Here the process was closer to film directing than traditional visualization: exploring space, finding points of view and deciding how it should be observed.
I worked on framing, focal lengths, compositing, and camera movements so that each shot had a function within the sequence, rather than just producing a collection of separate views of the building.
Some chambers present the architecture from the outside; others cross the courtyard, explore the relationship between interior and exterior or use architectural elements to progressively reveal the space.
03 — Visual Development in Canvas
With the planes defined, the next step was to move the project to Canvas to develop its visual language.
The intention was to build a coherent identity around a relatively contained palette: light stone, wood, glass, water and Mediterranean vegetation, combined with warm interior lighting and a blue hour outdoor atmosphere.
Through references and image generation, I developed this language on different points of view, seeking that everyone belonged to the same place and at the same time.
In this phase, architecture ceased to be just geometry and began to acquire materiality, light and atmosphere.
04 — From Image to Motion
Once the shots and their visual appearance were defined, I used the video models available in Canvas to transform the static imagery into small cinematic sequences.
The movements were deliberately kept contained: dolly, slider, and push-in, avoiding overly complex movements that competed with the architecture.
The goal was to use movement as an extension of the original composition, trying to preserve geometry, framing, reflections, materials, and visual continuity as the camera swept through the space.
In this way, each image ceased to function as an isolated result and began to be part of a sequence.
05 — Constructing the End Sequence
The result is a small piece of cinematographic Archviz built from different routes over the same space.
The sequence combines side views, approaches to the house, explorations of the courtyard, relationships between interior and exterior and a final protagonist shot that brings together a good part of the visual language developed during the process.
Rather than exhaustively representing the architectural project, the intention was to suggest how it might feel to walk through it.
From prompt to atmosphere
The entire workflow can be summarized in a very simple sequence:
Prompt → Space → Camera Direction → Visual Development → Movement → Atmosphere
What interests me most about Flow Studio in this context is precisely the combination of generative AI and creative direction.
It's not just a matter of asking for an image and accepting a result. The process makes it possible to generate an environment, explore it spatially, choose where to place the camera, decide how to photograph it, subsequently develop its visual identity and, finally, enter movement.
This switch is especially interesting for the Archviz: to go from generating imagery of architecture to directing a visual experience around architecture.
UMBRAL is a first exploration of how Flow Studio can be incorporated into architectural visualization and spatial storytelling workflows, using AI not only as a generation tool, but as part of a broader creative direction process.
Ultimately, the experiment revolves around a simple question:
What happens when we stop using AI only to imagine spaces and start using it to direct how those spaces are seen, walked, and experienced?
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.