Text-to-image prompt crafting in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding prompt specificity in text-to-image generation

Which of the following prompts is most likely to generate a clear and detailed image of a red apple on a wooden table?

A "A colorful scene"

B "A fruit on a surface"

C "A red apple on a wooden table with natural lighting"

D "An object"

💡 Hint

Think about how adding details helps the AI understand what to draw.

Option C provides specific details about the object, color, surface, and lighting, which helps the AI create a clear and accurate image.

❓ Predict Output

intermediate

Output of a prompt with conflicting style instructions

What is the most likely visual style of the image generated by this prompt?

"A futuristic cityscape, painted in the style of a watercolor and photorealistic"

A A purely photorealistic image with sharp details and no painterly effect

B A black and white sketch

C A flat cartoon-style image with no texture

D An image blending soft watercolor textures with realistic city details

💡 Hint

Consider how AI models combine multiple style keywords.

The AI tries to merge both 'watercolor' and 'photorealistic' styles, resulting in a soft but detailed image.

❓ Model Choice

advanced

Choosing the best model for detailed text-to-image generation

You want to generate high-resolution images with fine details from complex prompts. Which model is best suited?

A A large diffusion-based model trained on diverse high-res images

B A simple GAN trained on low-res images

C A text-only language model without image generation capability

D A small lightweight model optimized for speed but low detail

💡 Hint

Think about model size, training data, and architecture for image quality.

Diffusion models trained on high-res diverse data produce detailed, high-quality images from complex prompts.

❓ Hyperparameter

advanced

Effect of guidance scale on image generation

In text-to-image generation, increasing the guidance scale parameter typically causes the output to:

A Become more closely aligned with the prompt but less diverse

B Become more random and less related to the prompt

C Run faster but with lower image quality

D Generate images with fewer colors

💡 Hint

Guidance scale controls how strongly the model follows the prompt.

Higher guidance scale forces the model to stick closer to the prompt, reducing randomness and diversity.

🔧 Debug

expert

Diagnosing poor image output from a text-to-image model

You input the prompt "A sunny beach with palm trees" but the generated image is mostly dark and blurry. Which is the most likely cause?

A The prompt is too vague and lacks style or lighting details

B The model was run with a very low number of inference steps

C The model is trained only on indoor scenes

D The prompt contains unsupported special characters

💡 Hint

Consider how inference steps affect image clarity.

Too few inference steps cause the model to produce blurry, incomplete images.