Skip to main content

Prompt Framework

After covering various topics, we've gone through the key elements of the text prompt framework. But I realize it may be hard to remember, so this chapter summarizes the framework.

Official Framework​

First, Midjourney's official frameworkīŧš


It's simple, with four sections:

  1. Subject
  2. Details & Background
  3. Style, Medium, Artist
  4. Parameters

My Framework​

The official template will get you quite far as a beginner. Based on my experiments, I've refined it with:

  1. Type first - order impacts weighting
  2. Split their Middle/Last sections for better memorization

My template summarizes the framework in a table for selective use per scene:

What is the type?E.g. watercolor, illustration
What is the subject?Describe the main subject in detail - more helps. If unclear, Midjourney will randomly generate details. Two main categories:
  • Who: For people, describe gender, looks, expression, pose, clothing, etc.
  • What: For objects, describe what it is, material, color, etc.
What is the environment?Key elements are:
  • Background - usually either:
    • Solid color
    • Scenic backdrops like indoor/outdoor
  • Lighting - where is light falling on the subject? What type?
  • Weather - for outdoor shots, this adds realism.
Pick backgrounds suiting the subject, else may look odd. Leave blank for AI to generate if not too particular.
What is the composition?After deciding subject and environment, think about framing, e.g.:
  • What's the camera focus?
  • Which direction is the subject facing?
  • What's the subject-to-background ratio?
What is the shooting medium?Key factors:
  • Camera model
  • Film
  • Lens
  • Settings like exposure, aperture
What is the style?Remember with 4 W's:
  • When - What era's style?
  • Who - Whose style (person or organization)?
  • What - What art type/movement style?
  • Where - What country's style?
ParametersPhoto parameters like:
  • Sharpness
  • Aspect ratio
  • See parameter sections for details

Memorization Tips​

It looks long but resembles taking a photo (just optimized for weighting):

  1. See interesting subject/person (focus) in some environment (environment)
  2. Take out phone (camera), frame shot (composition), click shutter
  3. Post process if needed (style & parameters)