Continuing from the previous section: In-depth Midjourney: Common Issues and Optimization with Prompts (Part 1)
6. Using weights (double colons) to create special effects#
In this section, I will show you how to handle themes and scenes as examples, just because using weights to create themes and scenes is an interesting way to learn about weights.
🤖Creating Themes
- Find theme words that interest you, such as friendship, fear, safety, hope, war, romance, destruction, heroism, silence, heartbreak, vanity, youth, futurism, robots, space opera, mystery, etc.
- Write your prompt with the theme words and measure it based on how much you want the theme to be present in your visual (recommended values are 1-3), then write your prompt as usual.
- The theme will affect your prompt. Take the following prompt as an example, and then create your own prompt.
Sentience::3 This is a picture of a chaotic mess of external wiring and plumbing on the walls of a brutalist architecture building in a busy city evening atmosphere. The colors are mainly midnight-blue and oak-brown. --ar 5:9
Sentience::3 This is a picture of a chaotic mess of external wiring and plumbing on the walls of a brutalist architecture building in a busy city evening atmosphere. The colors are mainly midnight-blue and oak-brown. --ar 5:9
(Here, the emphasis is on the Sentience weight of 3)
🌳Creating Settings/Scenes/Backgrounds
- Imagine a background or environment and describe it in 3-5 words, such as "midnight forest clearing" or "cliffs overlooking the sea."
- Describe your prompt with the description and weight it to help it compete with the rest of your prompt (recommended values are 1-3).
- The prompt acts like a "stage" where the weighted parts are more like "main characters" and the rest of your prompt will be presented on it. Try the following prompt as an example, and then create your own prompt.
Twilight, vintage coffee house::1.5 Hipster female with blue eyes and long flowing brown hair poses wearing a floral tank and jeans, photorealistic edge-to-edge print in the style of Joe Ranft, Tom Ruegger, John Canemaker, Atelier Lulua --ar 5:7
Twilight, vintage coffee house::1.5 Hipster female with blue eyes and long flowing brown hair poses wearing a floral tank and jeans, photorealistic edge-to-edge print in the style of Joe Ranft, Tom Ruegger, John Canemaker, Atelier Lulua --ar 5:7
(Here, Twilight, vintage coffee house is given a weight of 1.5)
🧮Common Issues
Midjourney defines things based on their main standard features. Deviating from the standard features means Midjourney doesn't "recognize" it. To make Midjourney "recognize" non-standard things, you need to transform them into standard things. For example, you can't get "a woman with green skin," but you can get "a female humanoid creature with green skin" because for Midjourney, "woman" can't have green skin like a human prototype, but a humanoid creature based on a non-human prototype can have green skin.
Another good approach is to use negative weights. You may need to use negative weights for non-standard things. For example, if you're trying to provide "non-standard" characteristics for something, you can try assigning negative weights to the characteristics you don't want to see.
In the following example, we want to give a cat a non-standard third eye. We allow Midjourney to add the third eye by replacing the "cat" prototype with "felinoid" and guide Midjourney to not focus too much on the "two eyes" by giving it a negative weight.
third-eye, felinoid::1 two eyes::-0.5
third-eye, felinoid::1 two eyes::-0.5
(two eyes is given a weight of -0.5, reducing its impact in the image)
7. Use sentence-like fragments instead of word lists.#
In --v 5, write sentence-like fragments instead of word lists. "A young woman walks across campus" is more effective than "woman, young, walking, campus."
"A young woman walks across campus" is better than "woman, young, walking, campus."
8. Styling Images.#
In --v 5, to generate content other than photographic images, you need to refer to art movements, artistic techniques, genres, media types, game titles, directors, artist names, influences, time periods, etc.
To invoke the aesthetic style of an image, try referencing two or more of the following:
- Art Movement: Identifying the art movement in the prompt will introduce its style and techniques. Examples include Impressionism, Surrealism, or Pop Art.
- Media Type: Identifying the medium of the image will determine its aesthetics. Examples include photography, illustration, comics, conceptual art, storyboards, sculpture, etc.
- Media Title: Identifying media influences will affect its appearance. For example, references to "Spirited Away" or "The Wizard of Oz" or "Sid Meier's Civilization" or the video game "Duel."
- Artist Name: Referencing a specific artist's name or work will loosely invoke their unique style. Examples include Vincent van Gogh, Frida Kahlo, or Banksy.
- Technique: Referring to techniques will add that style to the image. Examples include impasto, pencil sketch, watercolor, or digital art.
- Time Period: Identifying the historical background of the image will invoke its aesthetics. For example, images from the Renaissance period, Baroque period, or modernist period.
- Geographic Location: Referring to regions and countries will influence the style. Examples include Japanese ukiyo-e prints, African tribal art, or American abstract expressionism.
9. Image Aspect Ratio#
Don't want a square? Add the parameter --ar X or --aspect X or even --h X --w Y at the end of the prompt. (These three methods express the same thing in different ways.)
The following aspect ratios can be added at the end of the prompt:
Square
--ar 1:1
Portrait
--ar 1:2 --ar 2:3 --ar 4:6 --ar 5:7 --ar 9:16
Landscape
--ar 2:1 --ar 3:2 --ar 6:4 --ar 7:5 --ar 16:9
Even though you can specify any aspect ratio you prefer, Midjourney doesn't actually produce output for all aspect ratios. Only certain resolution values are allowed, and Midjourney will choose the closest one to the aspect ratio you specified. For example, if you specify --ar 850:1100, you will get the following results:
Initial grid: 4:5 (256x320)
Default Upscale: 4:5 (1024x1280)
Light Upscale: 4:5 (1024x1280)
Max Upscale: 11:14 (1408x1792)
Beta Upscale: 7:9 (1792x2304)
Best and closest Midjourney aspect ratios for the attached image:
10. Optimizing with Parameters: Style, Chaos, Quality Parameters#
To optimize images with parameters, refer to the official documentation for this section:
Style
https://docs.midjourney.com/docs/stylize
Chaos
https://docs.midjourney.com/docs/chaos
Quality
https://docs.midjourney.com/docs/quality