Exercise 2: Your First AI Image
Goal of This Exercise
You will generate your first AI image and experience the difference between a simple and a detailed image prompt. By the end, you will understand the basics of image prompting.
Tools for this exercise:
- ChatGPT (uses DALL-E, available in the free version)
- Bing Image Creator (free, accessible at bing.com/images/create)
Step 1: The Simple Prompt
Start with a minimal prompt:
Prompt: "A dog"
Enter this prompt in ChatGPT or Bing Image Creator. The result will be a dog -- but a generic, random dog. Maybe a Labrador in front of a white background, maybe an illustration. The AI decides everything that you do not specify.
Observe:
- What breed is the dog?
- What style was chosen? (Photo, illustration, cartoon?)
- What is the background like?
- What is the lighting like?
Step 2: The Detailed Prompt
Now we add details:
Prompt: "A Golden Retriever, watercolor style, warm colors, in an autumn forest, soft light through the trees"
Compare the result with Step 1. The difference is usually dramatic. You have now specified:
- Subject: Golden Retriever (not just any dog)
- Style: Watercolor (not a photo)
- Colors: Warm (orange, yellow, brown)
- Setting: Autumn forest
- Lighting: Soft, through trees
Step 3: Try Style Modifiers
Take the same subject and change only the style. Try these variants:
"A Golden Retriever, photorealistic, studio photography, white background, professional lighting"
"A Golden Retriever, pixel art, retro game style, 16-bit graphics"
"A Golden Retriever, oil painting in the style of the Renaissance, dramatic lighting, golden frame"
"A Golden Retriever, minimalist line art, black on white, simple lines"
The same dog every time -- but completely different results. This shows how strongly the style modifier affects the image.
The Art of Image Prompting
Remember this formula for good image prompts:
Subject + Style + Lighting + Composition
| Element | Examples |
|---|---|
| Subject | Person, animal, landscape, object, scene |
| Style | Photorealistic, watercolor, pixel art, oil painting, cartoon |
| Lighting | Studio flash, golden hour, neon light, soft daylight |
| Composition | Close-up, bird's eye view, wide angle, symmetry |
!Biases in AI images
Generated images can reflect biases -- pay attention to diversity. For example, if you prompt "a doctor," many AI tools automatically generate a man. If you enter "a nurse," a woman often appears. Be aware of these biases and actively counter them by being specific in your prompts.
Bonus Task
Create an image that you could actually use -- for example:
- A profile picture for social media (Note: label it as AI-generated!)
- An illustration for a presentation
- A desktop wallpaper
You receive an AI image that does not match your expectations. What is the best next step?
Reflect
You have experienced how strongly the level of detail in your prompt affects the result. A single word ("watercolor" vs. "photorealistic") can change the entire image. The ability to write good image prompts is a new creative skill -- the more you practice, the better your results will become.
Next up is Exercise 3: Understanding a simple AI automation.