Visual Prompting 101: Thinking Like an AI

Prompt Engineering is dead; Visual Prompting is alive. In the ChatGPT era, you learned to say "Act as a marketing expert." In the Whisk era, you need to learn to find images. You need to find a "minimalist composition but complex lighting" image to tell the model "I want this feeling."

What is Visual Prompting?

The core logic of Visual Prompting is: Image is Instruction. An image contains tens of thousands of times more information than text.

• Structure Prompt: Use stick figures, line drawings, or 3D gray models to control the skeleton of the image.
• Aesthetic Prompt: Use movie stills, oil paintings, or photography to control the skin of the image.
• Semantic Prompt: Use real object photos to define the subject of the image.

Build Your Visual Library

An excellent WhiskPrompt user usually maintains a massive asset library (Mood Board) like a veteran film director:

1. Light Library

Collect various lighting effect images.

• Tyndall Effect
• Neon Noir
• Rembrandt Lighting When you need this light, just drag an image from this folder.

2. Texture Library

• Rusted metal
• Rough paper
• Transparent glass

3. Layout Library

• Minimalist negative space
• Golden spiral
• Symmetrical composition

Don't Let AI "Guess"

Previously, we wrote "masterpiece, best quality, 8k," which was essentially begging AI: "Please, draw it better." Now in Whisk, we directly give it a masterpiece: "Draw according to this standard."

This is what AI should be like: It is your hand, and you are the brain.