Parameters

Stable Diffusion

Reverse diffusion works by subtracting the predicted noise from the image successively.

Weighting Syntax

Stable Diffusion supports both parentheses and numerical weights for emphasis:

• Emphasis:

  • (keyword) increases emphasis slightly.

  • ((keyword)) increases it more significantly.

  • (((keyword))) or additional parentheses for even stronger emphasis.

• De-emphasis:

  • [keyword] reduces the importance of the keyword.

• Numerical Weights:

  • Use keyword:weight to directly set the weight.

  • Example: a cat:1.5, a dog:0.5 gives more prominence to "cat" than "dog."

  • Typical weights range from 0.0 (ignore) to 2.0 (high emphasis).

Combining Weighted Terms

Stable Diffusion allows complex prompts with weighted phrases:

• Example Prompt:

  A beautiful ((sunset over the ocean)):1.5, [a city skyline]:0.7, colorful skies:1.2

  • This prioritizes the sunset while de-emphasizing the city skyline.

Impact of Weighting

Weights directly influence the model’s focus, but the results depend on:

• Model Training Data: Some concepts or terms may already be prominent or weak due to the dataset.

• Prompt Complexity: Overloading the prompt with high-emphasis terms can dilute results.

• Inherent Ambiguity: The model interprets prompts probabilistically, so multiple runs might slightly differ.

Using AND/OR for Advanced Control

You can separate concepts using AND for better multi-object composition:

• Example:

  A futuristic city:1.2 AND a dragon flying in the sky:0.8

  • This creates a balance between the city and dragon elements

Examples

• Detailed Portrait:

  A ((realistic portrait)) of a woman with blue eyes:1.3, wearing a red scarf:1.0 --negative (blurry), (cartoonish)

• Fantasy Scene:

  A magical forest with glowing mushrooms:1.5 AND a fairy with delicate wings:1.2 --negative (dark shadows), (low resolution)

MidJourney

In MidJourney, weighting syntax allows users to emphasize or de-emphasize certain elements in their image prompts. This can guide the AI on which parts of the prompt are more important or should have a greater influence on the final output.

Weighting

What

Double Colons (::):

• Place :: followed by a numerical weight after a specific term or phrase.

• Add a space after the numerical weight.

• Example: a golden retriever::2 sitting in a field of flowers::0.5

  • The golden retriever is emphasized with weight 2 (more influence).

  • The field of flowers is de-emphasized with weight 0.5 (less influence).

Why

Weighting allows you to emphasize or de-emphasize specific parts of your prompt by assigning a numerical weight to them. This helps you guide the AI to prioritize certain elements over others in the image composition.

How

Weights are specified using a :: followed by a number. The number determines how much influence the preceding word or phrase has on the output.

Example

term1::weight1 term2::weight2 term3::weight3

a futuristic city::2, a sunset::0.5, a flying car::1

Notes

• Default Weight: If no weight is specified, it defaults to 1.

• Higher Weight: A number greater than 1 (e.g., ::2, ::3) increases influence.

• Lower Weight: A number less than 1 (e.g., ::0.5) decreases influence.

• Zero Weight: A weight of ::0 effectively ignores that term.

• Avoid assigning very high weights (e.g., ::10) unless you want the image to focus exclusively on that element.

• MidJourney automatically normalizes the overall influence of weighted terms in a prompt. If one term is heavily weighted, other terms with default or lower weights may be downplayed.

Examples Prompts

Balanced

a dragon::1, a castle::1, a knight::1

All elements have equal influence.

Emphasized

a dragon::2, a castle::1, a knight::0.5

The dragon is the primary focus, the knight is de-emphasized.

Dominant Background

a sunset::3, a beach::1, palm trees::0.8

The sunset dominates, while the beach and palm trees are secondary.

Negative Parameter

a futuristic city::2 a flying car::1 --no skyscrapers

You can use --no with weights to exclude or downplay certain elements. For instance:

This tells MidJourney to focus on the city and car but avoid including skyscrapers.

Combining Image Input with Text Weights

How?

When combining image references and text prompts, the --iw (image weight) parameter adjusts the influence of the uploaded image(s) compared to the text. By default, --iw is 1.

Example

[image reference link] a desert landscape::2, futuristic architecture::1 --iw 0.5

The text elements are emphasized more than the image input due to the reduced --iw value.

Use Cases

• Focus Control: Highlight specific objects, styles, or themes.

  • Example: a giant robot::2 city skyline::0.5

• Detail Reduction: Downplay less critical aspects of a scene.

  • Example: a dense forest::2 a small lake::0.2

• Experimentation: Test different weights to fine-tune output.

---

_{© 2024 by SopakcoSauce. Except as otherwise noted, the content of this page is licensed under CC BY-NC-SA 4.0}