Glossary. Terms from visualization and AI.

The spatial representation of a building or room in software, from which plans, views and images are made. A 3D model can produce visualizations from any perspective.

Learn moreSee also: White model, CAD

The area in Vizera where you rotate an uploaded 3D model and pick the camera. The image is created from the chosen view, and the geometry is preserved.

Learn moreSee also: 3D model, Archicad plugin

A visualization that an AI image model generates from a source – a model view, a sketch or a photo – instead of computing it from light and materials. The image is ready in minutes; light and shadow are depicted, not physically computed.

Learn moreSee also: Diffusion model, Image-to-image

A shading effect that darkens corners, joints and tight gaps where less ambient light reaches. It gives models depth, often already in the white model.

See also: White model

The connection between Archicad and Vizera: send the view, sync the model with cameras. For Archicad 28 and 29, on Mac and Windows.

Learn moreSee also: 3D viewer, CAD

The visual representation of a planned building or room before it's built. It serves design, discussions with clients and authorities, competitions and marketing. Traditionally it's made with rendering software from a 3D model, today also with AI from a model, sketch or photo.

Learn moreSee also: Rendering, AI rendering

A way of working in which a building's 3D model also carries information: components, materials, quantities, costs. All planners work on the same model.

Learn moreSee also: CAD

The workspace in Vizera where variants sit side by side – for comparing and deciding with your team or clients.

See also: Material variant

Computer-aided design: software for drawing and constructing plans and 3D models, such as Archicad, Revit, Rhino or SketchUp.

Learn moreSee also: BIM (Building Information Modeling), 3D model

Rendering on servers on the internet instead of your own computer. The office doesn't need a powerful graphics card for it.

See also: Render farm

The images a firm uses to present its entry in an architecture competition. Usually a few carefully coordinated perspectives for the boards and the jury.

Learn moreSee also: Perspective

The part of AI that analyzes images: edges, surfaces, spaces, objects. Image models use it to capture the structure of a source.

See also: Image-to-image

The unit of use in Vizera. An image uses 5 to 25 Credits depending on model and resolution; subscription Credits apply per month, purchased Credit packs don't expire.

Learn more

A model trained on a specific firm's visual language. At Vizera, this is a separately commissioned service on written order, with purpose, data and deletion set out in the contract.

Learn moreSee also: Training data

A branch of machine learning using multi-layered neural networks that recognize patterns in large amounts of data. Image and language models are based on it.

See also: Generative AI

An image model that forms an image step by step from noise, guided by an input: a text, a sketch, a model view. Most current image generators are based on this principle.

See also: Generative AI, Prompt

People, vehicles, plants and furniture that bring a visualization to life and make the scale readable. Too much of it distracts from the design.

See also: Perspective

The AI extends an image beyond its edge, for example with more surroundings, sky or ground, matching what's there. Useful when a format needs to be wider or taller.

Learn moreSee also: Masking (inpainting)

The same building envelope in other materials or colors, with the same perspective and surroundings. The basis for deciding on the facade before a sample is made.

Learn moreSee also: Material variant

A floor plan becomes readable as an image: colored with floors and furniture, as a 3D image from above or as a white model. This helps clients who don't read plans.

Learn moreSee also: White model

AI systems that generate new content – text, images, videos – based on patterns from their training data. Image models for architecture today mostly work with diffusion.

See also: Diffusion model, Training data

Computing indirect light, meaning light that bounces from walls, floors and ceilings further into the room. Without it, interiors look dark and harsh.

See also: Ray tracing

A panoramic image with high dynamic range that serves as the sky and light source of a scene. It provides natural ambient light and reflections.

See also: Lighting mood

An AI method that transforms an existing image into a new one while keeping its structure, such as day to night or sketch to visualization. Most AI work on architectural images is based on this principle.

See also: Text-to-image, Masking (inpainting)

The moment a trained model processes an input and delivers a result, such as an image from your source. The duration of a generation is called inference time.

See also: Diffusion model

One pass in the design process: change, review, decide, continue. The faster a pass, the more variants fit into a phase.

See also: Material variant

The effect light gives an image: morning, noon, evening, blue hour or night. The same building looks different depending on the time of day, which is why firms often show several versions.

Learn moreSee also: Season

An area of the image is marked and regenerated only there – a facade, the sky, a piece of furniture. The rest of the image stays as it is.

Learn moreSee also: Extend image (outpainting), Post-processing

A collection of ready-made surfaces – wood, stone, plaster, metal – with textures and material values for rendering. Classic renderers need it maintained; in AI tools, you describe materials in words or with a reference image.

See also: Reference image, PBR (Physically Based Rendering)

The same design in different materials, such as wood, plaster or clinker brick, side by side for comparison.

Learn moreSee also: Facade variant

A collection of images, materials and colors that captures a project's design direction before the design work starts.

Learn moreSee also: Reference image

Image synthesis with neural networks instead of classic light computation. It delivers believable images quickly but doesn't compute light physically.

See also: AI rendering

A material model that describes surfaces through physical values such as roughness, metalness and reflection. That way a material responds believably under any light.

See also: Material library

The standpoint and angle from which a building or room is shown: eye level, bird's-eye view, interior view. The choice decides what an image tells.

See also: Entourage

A planned building is inserted into a photo of the site to show its effect in context – for the building permit application, public consultation or neighbors.

Learn moreSee also: Perspective

An image barely distinguishable from a photograph: believable materials, light, shadow and surroundings. Common for competitions, marketing and presentations.

See also: Rendering

Editing a finished image: color correction, contrast, sky, entourage. Traditionally in an image editing program, with AI also through targeted changes to individual areas.

See also: Masking (inpainting)

The plain-language description you use to tell an AI what should be created or changed: material, light, surroundings. Specific details like "larch cladding, evening light" work better than general ones.

See also: Reference image

A method that traces the path of light rays through the scene to compute reflections, refractions and shadows physically correctly. The basis of most photorealistic renderers.

See also: Global illumination

Rendering fast enough that you can move through the model, usually as a plugin in the CAD program. It suits walkthroughs and coordination on the finished model.

Learn moreSee also: Rendering

An image that shows the AI how a material, mood or style should look, instead of just describing it. The result interprets the reference; it doesn't copy it.

See also: Prompt, Style transfer

A cluster of many computers that renders complex images and animations in parallel. Firms usually rent it as a service.

See also: Cloud rendering

Computing an image from a 3D model with camera, materials and light. In everyday use, the term also means the finished image. Classic rendering computes light physically; AI rendering generates the image with a trained model.

See also: AI rendering, Ray tracing

The number of pixels in an image, such as 1920 × 1080. Little is enough for screens and meetings; posters, site hoardings and competition boards need much more.

Learn moreSee also: Upscaling

The same design in spring, summer, fall or winter, with matching vegetation, snow and light. This shows how a building fits into its surroundings over the year.

Learn moreSee also: Lighting mood

A hand sketch becomes a visualization with material and light. Useful in the early phase, before a model exists.

Learn moreSee also: Image-to-image

The visual language of one image – colors, light, rendering style – is transferred to the content of another. In architecture, for example, so that several visualizations of a project look consistent.

See also: Reference image

An image is created from a text description alone, without a source. That's good for mood and ideas, but the building shown is invented, not your design.

Learn moreSee also: Image-to-image

The images and texts an AI model learns from. For firms, what matters is whether their own projects are used for it: Vizera does not train its own general models on customer data.

Learn moreSee also: Custom AI model

Enlarging an image to more pixels, with AI adding structures only hinted at in the small image. In Vizera up to four times, as the last step before print.

Learn moreSee also: Resolution

An empty room is furnished in the image, for example for a property listing. The furniture is a suggestion in the image; in the listing, the image should be labeled as virtually furnished.

Learn moreSee also: Masking (inpainting)

A building or room model without materials, only in white and gray. It shows volume, proportion and layout and works well as a source image for visualization.

Learn moreSee also: 3D model, Ambient occlusion