Unit IVComputer Studio - II

Visualization & Rendering

Turning a model into an image — with light, a camera, and now AI.

≈ 45 min + studio task

Rendering is the simulation of light that turns a model into an image. Learn the engines — fast rasterization, photoreal ray and path tracing, and the global illumination that makes a room glow — and the three levers that decide whether an image convinces: materials, lighting and the camera. Then the newest layer: AI visualization with MidJourney, Stable Diffusion and Adobe Firefly — what each does, and where it must not be trusted.

Learning objectives

By the end of this lesson, you will be able to — mapped to the course outcomes for Design of Structures I:

CO4 · Understand

Explain rendering — rasterization, ray/path tracing and global illumination.

CO4 · Apply

Set materials, lighting (natural, HDRI, artificial) and the camera for a convincing image.

CO4 · Apply

Export and convert images correctly for screen and for print.

CO6 · Understand

Describe how AI tools (MidJourney, Stable Diffusion, Firefly) are used — and their limits — in arch-viz.

Light, materials, camera

How rendering works

Rasterization is fast; ray and path tracing are photoreal; global illumination is the bounced light that sells realism. Materials and lighting — natural sun/sky or an HDRI, plus artificial lights — and a two-point camera do the rest.^{[1, 2, 3]}

DiagramLight types for a scene: a directional sun, a point light, a spot cone, an area light, and an HDRI dome for image-based lighting

DiagramThe architectural camera — two-point perspective keeps verticals vertical, with field of view and depth of field

Simulating light

Rendering turns a 3D scene into a 2D image by computing light. Rasterization projects polygons forward onto the screen — very fast, used for real-time. Ray tracing shoots rays from the camera into the scene; path tracing follows many bouncing rays to produce global illumination, soft shadows, reflections and caustics — the basis of photoreal offline renders.^{[1, 2]}

Interactive

A render builds up in passes

Toggle each pass and watch the image assemble — wireframe, clay, materials, lighting, shadows, reflections, post. Notice that most of the realism arrives with materials and lighting, not with any single quality setting.^{[2, 3]}

Render build-up · toggle the passes

Switch each pass on in turn and watch the image assemble. Most of the realism arrives with Materials and Lighting — not from any single quality setting.

DiagramA render building up in passes from wireframe through clay, materials, lighting, shadows, reflections to post-processing

MidJourney · Stable Diffusion · Firefly

AI tools for architectural visualization

MidJourney invents from text (ideation); Stable Diffusion with ControlNet follows your geometry from a depth or line export; Adobe Firefly is built into Photoshop for post-production. All of it is powerful — and none of it is measured documentation.^{[5, 6, 7, 8]}

Text-to-image ideation

MidJourney is a subscription text-to-image tool (web and Discord), prized for polished, atmospheric concept images and mood boards. By default it invents a building from your words — it does not keep your modelled geometry — so it is an ideation tool, not a way to render your actual design.^[5]

The contrasts

At a glance

Aspect	One	The other
Rasterization vs ray tracing	Raster: fast forward-projection (real-time)	Ray/path tracing: physical light (photoreal)
Real-time vs offline	Real-time: instant preview (Enscape, D5)	Offline: slower photoreal frames (V-Ray, Corona)
Natural vs artificial light	Sun/sky/HDRI — sets time-of-day and mood	Point/spot/area — controlled interior light
AI ideation vs documentation	AI image: evocative, dimensionally unreliable	Drawing: scaled, accurate, code-checkable
MidJourney vs Stable Diffusion	MidJourney: invents from text	SD + ControlNet: keeps your geometry

Vocabulary

Key terms

Rendering

Generating a 2D image from a 3D scene by simulating light.

Rasterization

Fast forward-projection of polygons onto the screen (real-time).

Ray / path tracing

Tracing rays (and many bounces) from the camera for realistic light.

Global illumination (GI)

Indirect, bounced light — the biggest contributor to realism.

Ambient occlusion (AO)

Soft contact-shadows in corners and where surfaces meet.

HDRI

A 360° high-dynamic-range image used to light a scene (image-based lighting).

Two-point perspective

An arch-viz camera that keeps vertical lines vertical.

ControlNet

A Stable Diffusion add-on that conditions output on depth/edges to preserve geometry.

Generative Fill

Adobe Firefly/Photoshop feature that adds, removes or extends image content from a prompt.

Apply it

Studio task

Render one view of your model twice — once with a midday sun, once at golden hour with the interior lights on — keeping everything else the same. Write two lines on how lighting alone changed the mood. Then export one for screen and one for a 300 DPI print, and note the format and colour-space difference.

Check your understanding

Self-assessment

1. Which technique follows many bouncing rays from the camera to produce global illumination?

2. For a high-quality CMYK print of a render, the best choice is —

3. Which Stable Diffusion feature turns a SketchUp depth/line export into a render while keeping the geometry?

In a nutshell

Recap

Rendering simulates light: rasterization is fast, ray/path tracing is photoreal, and global illumination makes scenes glow.

Materials, lighting (sun/sky/HDRI/artificial) and the camera (two-point perspective, lens, DoF) decide whether an image convinces.

Export correctly — TIFF/PNG at 300 DPI in CMYK for print, JPEG/PNG for screen.

AI tools — MidJourney (ideation), Stable Diffusion + ControlNet (geometry-faithful), Firefly (post-production) — are powerful but never measured documentation.

The evidence

References & further reading

[1]Rasterization vs ray tracing — NVIDIA Developer / GPU Gems. https://developer.nvidia.com/gpugems/gpugems2/part-v-image-oriented-computing/chapter-38-high-quality-global-illumination
[2]What is path tracing / global illumination — NVIDIA. https://blogs.nvidia.com/blog/what-is-path-tracing/
[3]PBR materials and HDRI image-based lighting — Chaos. https://blog.chaos.com/what-is-pbr-physically-based-rendering-a-complete-guide
[4]Choosing image file formats for print (TIFF/PNG/JPEG, DPI, CMYK) — CreativePro. https://creativepro.com/how-to-choose-the-right-image-file-format-for-print/
[5]Midjourney — plans and usage (subscription text-to-image). https://docs.midjourney.com/hc/en-us/articles/27870484040333-Comparing-Midjourney-Plans
[6]Stable Diffusion + ControlNet for architectural rendering — Stable Diffusion Art. https://stable-diffusion-art.com/controlnet/
[7]Adobe Firefly — generative-AI approach (training data, indemnification). https://www.adobe.com/ai/overview/firefly/gen-ai-approach.html
[8]AI renderers for architecture — capabilities and limits (Chaos / Veras). https://blog.chaos.com/best-ai-rendering-tools-for-architects-compared