Quiz: Image Generation
Test your knowledge about diffusion models, image generation platforms, and prompting techniques.
30 questions - Click on each question to reveal the correct answer.
Question 1
What is the core concept behind diffusion models?
A) Copying existing images
B) Learning to reverse a noise-adding process
C) Combining multiple images
D) Pixel-by-pixel drawing
Show Answer
Correct Answer: B
Diffusion models learn to generate images by reversing a noise-adding process—starting with random noise and gradually removing it to reveal an image.
Question 2
Which platform is known for producing the most artistic and aesthetic images?
A) DALL-E 3
B) Midjourney
C) Stable Diffusion
D) Adobe Firefly
Show Answer
Correct Answer: B
Midjourney is renowned for its exceptional aesthetic quality and artistic output, with a distinctive style that many find visually stunning.
Question 3
What does CFG (Classifier-Free Guidance) scale control?
A) Image resolution
B) How closely the output follows the prompt
C) Generation speed
D) File size
Show Answer
Correct Answer: B
CFG scale controls how closely the generated image adheres to the text prompt. Higher values mean more literal interpretation; lower values allow more creative freedom.
Question 4
Which image generation platform is open source?
A) DALL-E 3
B) Midjourney
C) Stable Diffusion
D) Adobe Firefly
Show Answer
Correct Answer: C
Stable Diffusion is open source, allowing anyone to download, run, and modify the model freely.
Question 5
What is a "negative prompt"?
A) A poorly written prompt
B) Instructions for what NOT to include in the image
C) A prompt that generates dark images
D) A prompt that fails
Show Answer
Correct Answer: B
A negative prompt specifies elements you want to avoid in the generated image, like "blurry, bad anatomy, watermark."
Question 6
What does the "seed" parameter do?
A) Plants a digital flower
B) Determines the random starting point
C) Sets the image quality
D) Chooses the style
Show Answer
Correct Answer: B
The seed determines the random starting point for generation. Using the same seed with the same prompt produces the same image.
Question 7
Which company created DALL-E?
A) Google
B) Stability AI
C) OpenAI
D) Midjourney Inc.
Show Answer
Correct Answer: C
OpenAI created DALL-E, including the current DALL-E 3 version integrated into ChatGPT.
Question 8
What is "latent diffusion"?
A) Slow diffusion
B) Diffusion in a compressed representation space
C) Hidden diffusion
D) Delayed diffusion
Show Answer
Correct Answer: B
Latent diffusion works in a compressed representation (latent space), making it much more efficient than working directly with full-resolution pixels.
Question 9
Where do you access Midjourney?
A) Website
B) Mobile app
C) Discord
D) ChatGPT
Show Answer
Correct Answer: C
Midjourney is accessed through Discord using the /imagine command in their server.
Question 10
What is DALL-E 3's main strength?
A) Artistic style
B) Accurate prompt following
C) Running locally
D) Lowest cost
Show Answer
Correct Answer: B
DALL-E 3 excels at accurately following complex prompts and instructions, better than most competitors.
Question 11
What hardware is typically needed to run Stable Diffusion locally?
A) Any computer
B) A GPU with at least 6GB VRAM
C) A quantum computer
D) Only cloud servers
Show Answer
Correct Answer: B
Stable Diffusion requires a GPU with at least 6GB VRAM, though 12GB+ is recommended for better performance.
Question 12
What does the --ar parameter do in Midjourney?
A) Sets artist style
B) Sets aspect ratio
C) Adds artistic effects
D) Archives the image
Show Answer
Correct Answer: B
The --ar parameter sets the aspect ratio, like --ar 16:9 for widescreen or --ar 1:1 for square.
Question 13
Which component should typically come first in an image prompt?
A) Quality modifiers
B) Lighting description
C) Subject description
D) Style reference
Show Answer
Correct Answer: C
The subject should typically come first, as it's the most important element. Formula: Subject → Style → Lighting → Composition → Quality.
Question 14
What was the main limitation of GANs that diffusion models solved?
A) Speed
B) Training stability and mode collapse
C) Cost
D) Resolution
Show Answer
Correct Answer: B
GANs suffered from training instability and mode collapse (generating similar outputs). Diffusion models train more stably and produce more diverse results.
Question 15
What is SDXL?
A) A Midjourney feature
B) An advanced version of Stable Diffusion
C) A DALL-E model
D) An image format
Show Answer
Correct Answer: B
SDXL (Stable Diffusion XL) is an advanced version of Stable Diffusion that produces higher resolution and better quality images.
Question 16
Which lighting term creates a warm, romantic atmosphere?
A) Blue hour
B) Golden hour
C) Dramatic lighting
D) Neon lighting
Show Answer
Correct Answer: B
Golden hour refers to the warm, soft light shortly after sunrise or before sunset, creating a romantic atmosphere.
Question 17
What is the typical number of denoising steps in diffusion models?
A) 1-5
B) 20-50
C) 100-200
D) 500+
Show Answer
Correct Answer: B
Most diffusion models use 20-50 denoising steps. More steps can produce finer details but take longer.
Question 18
What does CLIP do in image generation?
A) Clips unwanted parts
B) Converts text to numerical representations
C) Compresses images
D) Connects to the internet
Show Answer
Correct Answer: B
CLIP (Contrastive Language-Image Pre-training) converts text prompts into numerical embeddings that guide the image generation process.
Question 19
Which platform is best for rendering text within images?
A) Midjourney
B) Stable Diffusion
C) DALL-E 3
D) All equally good
Show Answer
Correct Answer: C
DALL-E 3 is notably better than competitors at accurately rendering text within generated images.
Question 20
What's a common negative prompt to avoid?
A) "beautiful"
B) "blurry, bad anatomy, deformed"
C) "high quality"
D) "detailed"
Show Answer
Correct Answer: B
Common negative prompts include "blurry, bad anatomy, deformed, extra fingers, disfigured" to avoid quality issues.
Question 21
Why would you choose Stable Diffusion over Midjourney?
A) Better quality
B) Easier to use
C) Privacy and customization
D) More artistic
Show Answer
Correct Answer: C
Stable Diffusion is chosen for privacy (runs locally), customization (full control), and cost (free to run).
Question 22
What is prompt weighting?
A) Making prompts longer
B) Emphasizing or de-emphasizing certain concepts
C) Adding more keywords
D) Using weighted average of images
Show Answer
Correct Answer: B
Prompt weighting allows you to emphasize (increase weight) or de-emphasize (decrease weight) certain concepts in your prompt.
Question 23
What does "volumetric lighting" create?
A) Flat lighting
B) God rays and atmospheric effects
C) Dark shadows
D) Neon glow
Show Answer
Correct Answer: B
Volumetric lighting creates visible light rays (god rays) and atmospheric effects, adding depth and mood to images.
Question 24
What is the approximate cost of Midjourney's Standard plan?
A) $5/month
B) $10/month
C) $30/month
D) $100/month
Show Answer
Correct Answer: C
Midjourney's Standard plan costs approximately $30/month (as of 2024).
Question 25
What is a U-Net in diffusion models?
A) A user interface
B) The neural network that predicts noise
C) A file format
D) A universal network
Show Answer
Correct Answer: B
U-Net is the neural network architecture that predicts the noise to be removed at each step of the diffusion process.
Question 26
Which is NOT a valid composition term?
A) Rule of thirds
B) Bird's eye view
C) Color temperature
D) Dutch angle
Show Answer
Correct Answer: C
Color temperature relates to lighting (warm/cool), not composition. The others are composition/framing terms.
Question 27
What does "8K" in a prompt typically indicate?
A) File size
B) High resolution/detail request
C) Number of colors
D) Processing time
Show Answer
Correct Answer: B
"8K" is a quality modifier requesting high resolution and detail in the generated image.
Question 28
What is the VAE's role in latent diffusion?
A) Generates text
B) Encodes/decodes between image and latent space
C) Adds noise
D) Connects to database
Show Answer
Correct Answer: B
The VAE (Variational Autoencoder) compresses images to latent space for processing and decodes results back to full images.
Question 29
When should you use multiple image generation platforms?
A) Never
B) When different projects have different requirements
C) Only for professional work
D) Only when one fails
Show Answer
Correct Answer: B
Many professionals use multiple platforms because each excels at different things—DALL-E for accuracy, Midjourney for aesthetics, Stable Diffusion for control.
Question 30
What's the best approach when your prompt doesn't give desired results?
A) Give up
B) Use a completely different platform
C) Iterate and refine the prompt
D) Add more random words
Show Answer
Correct Answer: C
The best approach is to iterate and refine your prompt based on the results, adjusting elements that aren't working.
Results
| Score | Level |
|---|---|
| 27-30 | Excellent - Strong understanding of image generation |
| 21-26 | Good - Some review may be helpful |
| 15-20 | Average - Review the chapters again |
| Below 15 | Needs comprehensive review |