Skip to content

Visualisation

How visualisation works

How the images of your home are made, what the “AI visualisation” label means, how we check them before showing them, and what they do not show.

Visualisations and image signing are in a pilot and not yet part of the subscriptions.

Three kinds of image

A “Render” is drawn from the house's 3D model with Blender (Cycles), with the finishes and products of the BoQ at their true size. It carries the badge “Rendered from the house model”. Apart from noise reduction (Intel Open Image Denoise) it is not processed further.

An “AI visualisation” starts from such a render. An AI image model changes only the unpriced parts: the furniture, the decor and the light. The pixels of every priced finish and product are copied back from the render, so they show exactly the chosen product.

An “AI photo edit” starts from your photo of a real room. New finishes are placed only in the areas you confirm, at the scale you set. The light around them can be evened out by a model that runs only on our own servers in the EU.

The label

Every AI visualisation carries “AI визуализация · AI visualisation” in its top-right corner, and a second line saying what changed: “AI: мебели, декор и светлина”, or “AI: редактирана снимка на реално помещение” for a photo edit. The label is part of the image itself, so it stays on downloads and prints.

AI visualisations are signed with C2PA (a provenance record) and carry an invisible watermark (Adobe TrustMark). Both can be checked on the “Check an image” page. We sign with our own certificate, so outside checkers show an “unknown signer”. Check an image

The check before showing

Before we show an AI visualisation, we compare it automatically with the render from the model:

  • the image lines up with the render within 1 pixel;
  • the colours of the priced finishes match the stored colour (ΔE2000, 95th percentile at most 2);
  • the windows and doors are where they were (at least 95 % of their outlines, within 2 pixels), and the walls and structure at least 85 %;
  • the blend between AI pixels and priced pixels is at most 3 pixels wide.

If the check fails, we try once more with a lighter change (at most 2 attempts). If that fails too, we show the clean render from the model, explain why (“Why it was rejected”) and give the credit back.

Renders from the model are checked too: if a finish's colour differs from the stored one, the render is marked “Colour not verified”.

What the image does not show

The image is an illustration. The prices and quantities are in the BoQ, not in the picture. Furniture, decor and the view are not in the BoQ; added furniture is marked “Image only — not in the bill of quantities”.

When a product has no texture of its own, we show a generic material and mark it “(similar product)”.

Limits

Per month: the Home plan includes 20 renders and 5 AI visualisations, Professional 100 and 20, Business 300 and 60. A repeat render from the same camera with the same finishes does not count.

On top: a pack of 50 renders for €5.90 and a pack of 20 AI visualisations for €9.90 (incl. VAT).

If the platform's overall limit for AI images is reached for a while, or no AI model is available, we show the clean render and the credit is not lost.

Visualisation is being introduced step by step: for now it is open only to projects in the pilot.

Your data

AI visualisations are switched on per project (“AI images”) and are off by default.

We send outside AI models only renders of the house model, never your photos, and only to providers that keep no data. Photos of real rooms are processed only on our own servers in the EU.

If something is wrong

If an image is offensive or wrong, use “Report image” below it. The report reaches the project owner and us.