AI rendering in architectural presentations by Sedat Bayrak

Sedat Bayrak is a practicing architect who runs his own office and still models and draws himself. This video is a broader conversation about AI tools in architecture, from language models for research to image and animation tools, and most of it is spent on how his office uses Rendair on delivered projects. What makes it worth watching is the practice view: he is clear that a trained eye has to check everything the AI produces, and he explains how he talks to clients about it.

What you'll see

1. A short overview of the other AI tools he uses for research, graphics and animation, and why Rendair replaced earlier image tools for his architectural work.
2. 3D Model to Render on plain SketchUp exports of interiors, to test whether a separate visualization step could be skipped.
3. Image to Render on renders from Lumion and old V-Ray images from his office's early years, run at low, medium and high creativity.
4. Layering the best parts of several generations in Photoshop, then fixing people last, a step that matches today's Fix People.
5. A very old, low-resolution render run at high creativity to explore new themes, such as Mediterranean or modernist versions of the same scheme.
6. Prompts written with help from a chatbot that describes his image using terms from Rendair's keyword list.
7. Residential interiors from a finished project, with a close look at what the AI got right and what it misread.
8. Image to Video with camera motion presets.

Takeaways

• He found generic objects like sofas, tables, rugs and window frames came out very well, and he praised how lighting fell into corners and across ceilings. Plants, gravel and landscape also worked well.
• Some details failed. A special open-cell ceiling never came out as designed, wood came out too knotty, and the AI misread items such as a globe and a wooden pendant. Those needed several more runs or a patch from his own render.
• He feeds in renders from Lumion or similar tools rather than raw SketchUp exports, because they carry sun direction and light and shadow information.
• He warns that texture and style are hard to repeat months later after a client revision. His office tells clients up front that AI is part of the process and frames it as enhancing their own design.
• For video, he sticks to zoom-in or static camera moves. Pulling back makes the AI invent parts of a building that will actually be built.

Try it yourself

Run a Lumion or V-Ray render through Image to Render at a few creativity levels, then use Fix People and Upscale on the version you keep.

The video is in Turkish.

Recorded with an earlier version of Rendair, so some screens and app names look different today.