Renders and AI Images: Design Intent, Not Quantities

Renders sell designs. A good render helps an owner choose a facade colour or a kitchen layout in minutes. Since about 2023, AI image tools have made these pictures even faster and cheaper to make.

But a picture is not a drawing. This lesson is about knowing the difference.

Rendering tools

A render engine takes a 3D model and adds materials, light and camera angles to make a photo-like image or a walkthrough.

• Common tools: V-Ray, Enscape, D5 Render, Lumion, Twinmotion, Corona.
• Inputs: a SketchUp, 3ds Max, Rhino or Revit model.
• Outputs: .jpg and .png images, videos, panoramas.

A render is only as accurate as the model behind it. If the model is rough, the render is rough, even if it looks perfect.

AI image tools

AI image tools generate or restyle pictures from a text prompt, a sketch or a rough render. Some are general (Midjourney is one well-known example). Others sit inside design and render software.

They are useful for mood boards and quick options. But they have a basic limit: they invent detail. An AI tool may add a window, change a ceiling height, or widen a room to make the picture look better. It has no idea what the drawings say.

The concept deck problem

Interior and architecture offices often send a concept deck: a set of slides or a PDF with plans, views and pictures. A typical deck mixes three kinds of image:

1. CAD plan snippets copied from the drawings.
2. 3D model top views or perspective views from SketchUp or 3ds Max.
3. Photo-real perspectives, often made or reworked with AI tools.

These come from different stages and different people. They often disagree.

In the made-up example above, the plan says the bedroom is 3.60 m wide with one window and a wardrobe on the left. The model top view looks about 3.9 m wide with the wardrobe on the right. The AI perspective shows two windows and no wardrobe. None of them is "wrong" as a concept. But only one can be built, and you cannot tell which from the deck.

Why this matters for cost

Say a QS prices wardrobes from the perspective images. The plan shows a 2.4 m wardrobe. The model shows 2.7 m. At ₹1,800 per sq ft of front area and 7 ft height (illustrative), the difference is:

• 2.4 m = 7.87 ft, so 7.87 x 7 = 55 sq ft, about ₹99,200.
• 2.7 m = 8.86 ft, so 8.86 x 7 = 62 sq ft, about ₹1,11,600.

That is about ₹12,000 on one wardrobe, from reading the wrong picture. In a whole flat, the gap can run into lakhs.

The rule

Treat renders, AI images and concept decks as design intent. Use them to understand what the owner wants. For quantities and prices:

• Ask which view governs.
• Measure from a dimensioned plan, elevation or shop drawing.
• If none exists, write your assumptions next to every quantity and get them confirmed before you quote.

What comes next

Every design hand-off passes a file from one team to the next. The next lesson follows those files from brief to handover.