Skip to main content

Prompting Guide

Note: Special thanks to contrinsan for providing this writeup!

Overview​

The key to prompting is to be specific without going over the token limit - include precise details about the scene, subject, style, and camera movements to ensure the AI accurately interprets your vision. Use the following guidelines to help compose a clear and coherent prompt that will get the results you want.

  • Subject: Clearly define the main character or object.
  • Scene/Environment: Describe the setting, including location, time of day, or weather.
  • Action: Specify what the subject is doing, using dynamic verbs for motion.
  • Style: Indicate the visual style (e.g., cinematic, anime, photorealistic).
  • Atmosphere/Mood: Convey the emotional tone (e.g., serene, dramatic, eerie).
  • Camera Movements: Include specific camera instructions (e.g., zoom in, pan left, tracking shot).
  • Lighting: Describe lighting conditions (e.g., soft sunlight, neon glow).
  • Shot Size: Specify framing (e.g., wide shot, close-up, medium shot).
  • Verbs: Employ dynamic verbs like "running," "zooming," or "tilting" to guide the model in creating motion.
  • Incorporate Metadata: Add tags like "hdr," "360-degree," or "fisheye" for specific image types.
  • Model-Specific Considerations: Tailor prompts to the strengths of the model. For Hunyuan, use highly descriptive prompts with cinematic terminology (e.g., "wide-angle view," "lens flare") and detailed environmental cues like weather or time of day.
  • Prompt Length: Aim for 60–100 words to provide sufficient context without overwhelming the model.
  • Logical Sequence: Organize the prompt to paint a clear picture, starting with the subject and scene, followed by actions, camera work, and stylistic details.
  • Avoid Overloading: Balance creativity with clarity to prevent confusing the AI.

Example​

Prompt​

A lone samurai in traditional armor stands on a misty cliffside overlooking a lush valley at dawn.
He unsheathes his katana, performing a slow, deliberate sword dance.
The scene is cinematic, with a serene yet intense atmosphere.
The camera starts with a wide-angle shot, slowly zooming in to a medium shot of the samurai’s focused expression.
Soft golden sunlight filters through the mist, casting gentle shadows.
The video is high-definition, emphasizing realistic textures and fluid motion.

Explanation​

  • Subject: "A lone samurai in traditional armor" clearly defines the main character.
  • Scene/Environment: "Stands on a misty cliffside overlooking a lush valley at dawn" sets a vivid location and time of day.
  • Action: "Unsheathes his katana, performing a slow, deliberate sword dance" uses dynamic verbs to describe the subject’s movement.
  • Style: "Cinematic" specifies a professional, film-like visual style.
  • Atmosphere/Mood: "Serene yet intense" conveys the emotional tone.
  • Camera Movements: "The camera starts with a wide-angle shot, slowly zooming in to a medium shot" provides specific camera instructions.
  • Lighting: "Soft golden sunlight filters through the mist, casting gentle shadows" details the lighting conditions.
  • Shot Size: "Wide-angle shot" and "medium shot" clarify the framing.
  • Metadata: "High-definition, emphasizing realistic textures and fluid motion" adds precision for quality output.
  • Word Count: The prompt is 73 words, fitting the recommended 60–100 word range for sufficient detail without overloading.