Prompt Engineering and Generative Model Examples

2025-11-20: Prompt engineering and examples are being updated in parallel, covering textbook-level illustration prompts and Gemini 3 preference guidelines. A set of high school physics illustration prompts for 'projectile motion trajectory and decomposition' is provided, including image type, horizontal composition, three-part layout, and visual center requirements. The illustration prompts detail the left, center, and right elements, coordinate system annotations, velocity decomposition, displacement decomposition, and the positioning and font specifications for formula boxes. The prompts specify color values (e.g., D32F2F, 1976D2, 4CAF50) and line thickness (0.8pt/0.6pt/0.4pt). It mentions that Gemini 3 prefers direct and logical prompts, avoiding lengthy preachy texts, and includes core principles and three key practices of prompt engineering (citing blogs). Example prompts achieve a Polaroid effect: after generation, users need to 'shake it' and wait for development, automatically analyzing expressions and generating warm text; images can be clicked to directly input prompts for re-editing, with prompts stored in the ALT field for reuse.
Key insights
Refined Prompt Examples Yield Textbook-Level Teaching Illustrations
A prompt targeting high school projectile motion demonstrates how precise constraints (image type, three-part layout, color and font size, coordinates and annotations, trajectory and component identification, etc.) can generate textbook-style diagrams. The prompt specifies background color values, accent colors, line thickness, and font sizes, detailing the elements in the left, center, and right regions, as well as the positions of formulas, aiming to ensure that the output images are scientific, intuitive, and easy for students to understand. This example illustrates that specifying layout and visual standards in prompt engineering can control the normativity and teachability of the generated results.
Gemini 3 Prefers Concise, Direct, and Logical Prompt Structures
Best practices in prompt engineering indicate that Gemini 3 prefers direct and logical prompts rather than lengthy preachy texts. Related blogs and summaries outline the core principles of prompts and three key practices, emphasizing the impact of instruction precision and structured expression on generation quality. This preference should be prioritized when crafting prompts to enhance the model's response accuracy for complex tasks, such as multi-step image editing or information-dense explanations.
Creative Prompt Examples Achieve Interactivity and Re-editing Features (Polaroid Effect)
Community-sourced examples showcase a Polaroid effect: generated photos require 'shaking' and waiting a few seconds for development, during which the model analyzes the expression at the time of capture and automatically generates warm text on the photo; images support direct prompt input for re-editing upon clicking, with prompts stored in the ALT field for reuse and sharing. This example also demonstrates the potential of prompts to enhance interactivity and user experience.