AI Image Generation Prompt Engineering

Professional Techniques: Glossary

Edition: January 2026

Addendum: Glossary of Terms, Acronyms, and Abbreviations

This addendum provides definitions for key terms, acronyms, and abbreviations used throughout the textbook. Each entry includes a concise definition and an example of use in the context of AI image generation prompt engineering. Entries are listed alphabetically for ease of reference.

• –ar (Aspect Ratio): A model-specific parameter used in tools like Midjourney or Stable Diffusion to define the width-to-height ratio of the generated image.
  Example: In a prompt, “–ar 16:9” ensures the output is widescreen, suitable for cinematic landscapes.
• Aspect Ratio: The proportional relationship between the width and height of an image, often specified to control composition.
  Example: Selecting “16:9” in CapCut’s UI for video-friendly social media visuals.
• CGI (Computer-Generated Imagery): Digital visuals created using computer software, often mimicking real-world appearances.
  Example: “Realistic CGI style” in a prompt to generate lifelike space scenes.
• Cinematic: A style evoking film aesthetics, including dramatic lighting, composition, and depth.
  Example: “Cinematic composition from a low-angle 16mm lens” to create epic, movie-like fantasy images.
• Composition: The arrangement of visual elements within an image frame, such as rule-of-thirds or symmetrical layouts.
  Example: “Symmetrical composition” in Nano Banana prompts for balanced infographics.
• Constraints: Explicit rules or limitations in a prompt to guide the AI and prevent unwanted outputs.
  Example: “No distortions, accurate anatomy” to ensure realistic human figures.
• Depth of Field: A photographic effect where only part of the image is in sharp focus, blurring foreground or background.
  Example: “Cinematic depth of field” in CapCut prompts to emphasize subjects in portraits.
• Diffusion Models: AI architectures (e.g., Stable Diffusion) that generate images by iteratively denoising random data.
  Example: Using Flux or SD family models for detailed artistic freedom in photorealism.
• Factual Grounding: Ensuring generated content aligns with real-world knowledge or data.
  Example: In Nano Banana, “Factual accuracy based on real historical data” for educational timelines.
• God Rays: Volumetric light beams piercing through atmosphere, creating dramatic effects.
  Example: “Dramatic volumetric god rays” in prompts for epic fantasy environments.
• HDR (High Dynamic Range): Imaging technique capturing a wide range of light intensities for more realistic contrasts.
  Example: “High dynamic range lighting” to enhance realism in outdoor scenes.
• Image-to-Image: A generation mode where an input image is transformed based on a prompt.
  Example: In CapCut, “Transform this photo into anime style” for style shifts.
• Infographic: A visual representation of information or data, often using charts, icons, and text.
  Example: Nano Banana prompts for “Clean vector infographic timeline” in branded assets.
• Inpainting/Outpainting: Editing techniques to fill in or extend specific image areas.
  Example: Flux/SD editing strength for refining generated portraits.
• Iteration: The process of refining prompts through successive modifications and generations.
  Example: “Iterate by changing only the lighting descriptor” in practice assignments.
• Midjourney: A diffusion-based AI image generator known for cinematic and artistic outputs.
  Example: Using “–stylize 750 –v 7” parameters for masterpiece-level renders.
• Nano Banana: Codename for Google’s Gemini advanced image generation models, emphasizing reasoning and text fidelity.
  Example: “Reason step-by-step about composition” in Pro version prompts.
• Negative Prompt: A list of elements to exclude from the generated image to avoid flaws.
  Example: “Blurry, lowres, deformed” in Stable Diffusion to improve quality.
• Photorealistic: A style mimicking real photographs with high detail and accuracy.
  Example: “Photorealistic product photography” for professional mockups.
• Prompt Engineering: The craft of designing effective text inputs to guide AI models in generating desired outputs.
  Example: Structuring prompts logically for optimal AI image results.
• Prompt Weight: A mechanism (e.g., (1.2)) to emphasize or de-emphasize elements in a prompt.
  Example: “Majestic ancient dragon (1.2)” to prioritize the subject’s detail.
• Reasoning: The AI’s step-by-step logical processing, prominent in models like Nano Banana Pro.
  Example: “Reason step-by-step about icon placement” for consistent designs.
• Rule-of-Thirds: A composition guideline dividing the frame into thirds for balanced placement of elements.
  Example: “Rule-of-thirds composition” in prompts for dynamic portraits.
• SD (Stable Diffusion): An open-source diffusion model family for text-to-image generation.
  Example: Medium-long detailed prompts for artistic photorealism.
• Specificity: The level of detail in descriptors to achieve precise AI outputs.
  Example: “Highly specific physical traits, age, ethnicity” in universal templates.
• –stylize: A Midjourney parameter controlling artistic abstraction level.
  Example: “–stylize 750” for highly stylized cinematic masterpieces.
• Text Rendering: The AI’s ability to generate clear, legible text within images.
  Example: “Maximum text legibility” in Nano Banana for infographics.
• UI (User Interface): The graphical elements through which users interact with software.
  Example: CapCut’s “UI style selector” for choosing anime or trending categories.
• –v (Version): A parameter specifying the model version in tools like Midjourney.
  Example: “–v 7” to access the latest features for improved outputs.
• Vector: A scalable graphic format using paths, ideal for clean designs.
  Example: “Clean vector infographic” in Nano Banana for diagrams.
• Volumetric Lighting: Light simulation accounting for atmospheric scattering and volume.
  Example: “Dramatic volumetric teal-pink lighting” in CapCut for moody scenes.