Gains:
- Ability to create strong visual demands layered with scene, view, light, material, scale element, style and climate constraints
- Ability to understand that artificial intelligence visuals are a communication tool and not a scaled technical document, and to transparently state that they are representative to the customer.
- Ability to inspect produced images for climate compatibility, scale and realistic density and correct exaggerated or inapplicable elements
There is a huge difference between telling a customer "there will be a shaded walking axis here" and showing a visual of what that axis looks like in the evening light. Rendering is a lifelike visualization of a design — making lines on paper “photo-like” with light, texture, and atmosphere. Text-to-image (artificial intelligence that produces images from text) is a tool that turns a recipe you write into a visual. In this unit, you will learn how to use AI in concept visual, moodboard, atmosphere sketch and fast rendering production; But you will learn why these images are a communication tool and not a scaled technical document.
Two different jobs of visual production
Separate the AI visual for two purposes. The first is atmosphere/inspiration: “how does this place feel?” Visual answer to the question — moodboard, concept illustration, reference atmosphere. One-to-one accuracy is not sought here; the goal is to convey emotion and direction. The second is suggestion visual: translating an existing sketch or CAD drawing into a realistic scene (converting to visual). Remember here too: the AI image is not to scale; Tree heights, distances and proportions are representative.
The critical distinction is this: AI imagery is great for saying “this is the feeling/atmosphere we are aiming for”; It is not meant to say "it will be done exactly to this extent, in this type, here". The second is the job of technical drawing (Unit 8).
Caution: Do not present the AI rendering to the customer saying "this is exactly what will be done". The image may contain elements that cannot be constructed in reality (a non-existent landscape, an impossible density of vegetation). Always keep in mind that "it is an atmosphere visual, it is not a definitive project".
How to set up a good visual prompt?
Text-to-image prompts are built layer by layer:
- Scene type: town square / streamside walk / roof garden.
- Perspective: eye level / bird's eye view / human perspective.
- Time and light: golden hour / morning fog / cloudy noon.
- Atmosphere and material: natural stone, wooden deck, herbaceous plant texture.
- People and scale: people walking, groups sitting (makes you feel the scale of the place).
- Style: photorealistic / watercolor sketch / architectural illustration.
- Constraints: plant texture suitable for the local climate, modest density.
As you add these layers, the output will be both beautiful and suitable for your purpose.
three mini cases
Case 1 — Fast atmosphere. An office produced 3 different atmosphere images (family during the day / social in the evening / sports in the morning) for a coastal promenade in half a day; In the past, these images would come from an external rendering company within a week and for a high fee. Three atmospheres were compared at the customer meeting, the direction became clear.
Case 2 — Wrong expectation (and its correction). A designer showed the AI render to the client as the "definitive view"; There was a density of flowers in the image that was impossible in reality. The customer was disappointed in the application. The office added a “representational atmosphere” watermark to all AI images from now on, and the expectation issues were gone.
Case 3 — From sketch to stage. There was only a rough plan sketch for a campus courtyard. The sketch was translated into a photorealistic eye-level scene by AI; The "living" version of the design was seen in the jury presentation. The image was not to scale, but it conveyed the idea powerfully; technical sheets were also presented.
Four copyable templates
1) Atmosphere image prompt:
A photorealistic landscape image: [streamside pedestrian promenade], eye level view, golden hour light, natural stone floor and wooden seating elements, herbaceous plant texture [suitable for temperate climate], people walking and sitting (for scale), calm and inviting atmosphere. Understated, realistic plant density.
2) Moodboard recipe:
Describe to me a moodboard concept for [arid climate, low water, shade-first schoolyard]: 6 image frames (material, shadow element, plant texture, seating, play, color palette) and a one-sentence description for each frame. Then turn them into visual prompts.
3) Style variation:
Describe the same scene (town square, eye level) in three different styles: 1) photorealistic rendering, 2) watercolor concept sketch, 3) architectural line illustration. We do not require separate images for each.
4) Visual verification check:
Evaluate this AI image for realism: is the vegetation density realistic, is the scale/person ratio correct, are there any elements that do not fit the climate, are there any impractical details? List the points that need to be fixed without showing it to the customer.
Weak prompt / Strong prompt
Weak prompt:
Make a beautiful garden image.
There is no stage, no lighting, no scale and no climate; The output would be clichéd and unfamiliar to the locale, possibly exaggerated.
Powerful prompt:
Photorealistic landscape rendering: a schoolyard suitable for a semi-arid climate, eye level view, morning light, shade pergola and broadleaf shadow texture, crushed stone and stabilized ground, children playing (for scale), herbaceous texture with low water consumption, realistic and understated density, warm and safe atmosphere.
Climate, light, material, the element of scale, and the constraint of “understated density” make the output seem both believable and feasible.
Image type selection table
Purpose
Suitable type
expectation of accuracy
note
conveying the emotion
atmosphere image
low
"Representative" grade is required
gather inspiration
mood board
low
reference intentional
present the idea
scene from sketch
medium
not to scale
compare decision
Style/atmosphere variation
medium
show side by side
technical explanation
NOT visual, sheet
high
CAD/scale drawing
Common mistakes
- Presenting the render as "final view". The AI image is not applicable to scale and one-to-one; Think “representational” note.
- Plant tissue foreign to the climate. If you do not specify climate in the prompt, the image may appear tropical/alien; Check cultural and ecological compatibility.
- Exaggerated intensity. AI images can produce clouds of flowers that cannot be dried in reality; Manage expectation.
- Stage without people. Without the scale element (human), space size cannot be read.
- Substituting the visual for the technical document. There should be scaled sheets next to the visual in the presentation.
In summary
AI visual generation is a powerful accelerator in landscape communication: reducing scene production from atmosphere, moodboard and sketch to hours. But these images are a communication tool, not a scaled technical document. A good prompt incorporates layers of scene, view, light, material, scale element, style and climate constraint. Check each image for authenticity, indicate to the customer that it is "representative" and present it with technical sheets.
Application task
Produce two different time/light (day and evening) images for a project with the "Atmosphere image prompt" template. Then check both for authenticity with the "Visual verification check" template and fix at least two points that go against the climate/scale.
checklist
- [ ] I specified scene, light, material, scale element and climate in the prompt.
- [ ] I checked the image for authenticity and climate compatibility.
- [ ] I fixed images containing exaggerated density/impossible elements.
- [ ] I pointed out to the client that the image was a "representative atmosphere".
- [ ] I added a scaled technical document next to the visual in the presentation.