You’ve mastered the basics of Midjourney. You can generate a “cyberpunk city” or a “photorealistic cat” with ease. But you’ve hit a wall. Your images are good, but they lack precise control, feel a bit generic, and you definitely can’t get the same character to appear twice. How do you get to the next level?
The secret isn’t just in what you ask for, but how you ask for it. While the core principles of clear communication are key to all AI (which we cover in our Ultimate Guide to Prompt Engineering), Midjourney has its own powerful and specific “language.”
Mastering a few of its advanced commands, or parameters, can be the difference between a random image and a true masterpiece. In this guide, we’ll cover 5 advanced Midjourney prompting techniques to give you precise control over your creations.
1. Blend Concepts with Multi-Prompting and Weights
One of the first challenges users face is getting Midjourney to understand the relationship between two different subjects.
The Problem: If you type a king holding a lion, Midjourney might create a “lion-king” hybrid, a lion wearing a crown, or a king with feline features. It blends the concepts.
The Solution: Use a multi-prompt by separating your concepts with a double colon (::). This tells Midjourney to consider each concept individually before blending them.
- Prompt:
/imagine prompt: a king:: holding a lion::
This simple change forces Midjourney to generate a distinct “king” and a distinct “lion” and then figure out how to combine them.
Take it a Step Further with Weights: You can assign relative importance to each part of your prompt by adding a number right after the double colon.
- Prompt:
/imagine prompt: a futuristic city::2 with a small::1 victorian house - What this does: This tells Midjourney to put twice as much emphasis on the “futuristic city” as the “victorian house,” ensuring the house is just a small element within the city, not a 50/50 blend.
2. Master Key Parameters for Precise Control
Parameters are commands you add to the end of your prompt to change the “rules” of the generation. While there are many, these three are essential for advanced control.
The “Chaos” Parameter: --chaos or --c
What it does: The --chaos parameter controls how varied and “surprising” your initial 4-image grid will be. It accepts a number from 0 to 100.
--c 0(Default): The images will be very similar and close to your prompt.--c 100(Max Chaos): The images will be wildly different, often using different styles, compositions, and subjects.- Why use it? When you’re brainstorming and want a wide range of ideas from a single prompt, cranking up the chaos (
--c 50) is a powerful way to discover unexpected creative directions.
The “Stylize” Parameter: --stylize or --s
What it does: The --stylize parameter controls how strongly Midjourney’s default “artistic” style is applied. It accepts a number from 0 to 1000 (in V6).
--s 0: Turns off the style, resulting in a more literal, “raw” interpretation of your prompt, which can be better for photorealism.--s 750(High): A very high-style, “artistic” and “opinionated” image.- Why use it? If your prompts are being over-stylized and you want a more literal photo, add
--s 0. If your image feels flat and you want more of that signature Midjourney flair, try--s 500.
The “Image Weight” Parameter: --iw
What it does: When you use an image prompt (dragging an image into the prompt bar), the --iw parameter tells Midjourney how much to respect your image versus your text. It accepts a number from 0 to 3 (in V6).
--iw 0: Ignore the image, only use the text prompt.--iw 1(Default): A good balance between the image and the text.--iw 3(Max): Pay very close attention to the source image, and less to the text.- Why use it? This is the key to controlling image-to-image generation. If you want to keep the composition of your source image but change the style, use a high image weight (
--iw 2or--iw 3) along with a text prompt describing the new style.
3. Save Time with Permutation Prompts {}
This is less about a single image and more about an advanced workflow. If you want to test multiple variations of one idea, you’d normally have to type the prompt over and over. Permutation prompts let you do it all at once.
What it does: By placing a comma-separated list of options inside curly braces {}, you tell Midjourney to create a separate job for each combination.
- Prompt:
/imagine prompt: a knight wearing {red, blue, green} armor - What this does: Midjourney will generate three separate jobs:
a knight wearing red armora knight wearing blue armora knight wearing green armor
You can stack them to test multiple variables at once:
- Prompt:
/imagine prompt: a {warrior, mage} in a {forest, castle} - What this does: This creates four jobs (warrior in forest, warrior in castle, mage in forest, mage in castle), letting you rapidly prototype your ideas.
RELATED POSTS
- The Ultimate Guide to Prompt Engineering: Getting Better Results from AI Tools
- The Top 3 Free AI Image Generators in 2025 (Ranked & Reviewed)
- Top AI Companies to Invest In (2025): A Guide for Investors
4. Get Consistent Style with Style References (--sref)
This is one of Midjourney’s most powerful new features. It solves the problem of “How do I get the vibe of this image without copying the content?”
What it does: The --sref parameter allows you to use one or more images as a “style reference.” Midjourney will ignore the content of the reference image and apply only its visual aesthetic (colors, mood, lighting, texture) to your new prompt.
How to use it:
- Find an image (or images) with a style you love. Upload it to Discord to get a URL, or find a URL online.
- Write your prompt.
- Add the parameter at the end:
--sref https://community.adobe.com/t5/stock-contributors-discussions/can-i-use-quot-style-references-quot-based-on-my-own-images/td-p/14634604
- Prompt:
/imagine prompt: a portrait of a stoic king --sref https://.../image-of-a-dark-oil-painting.jpg - What this does: It generates a new portrait of a king, but it will be rendered in the same dark, moody, oil-painting style as your reference image. This is a game-changer for creating a consistent brand aesthetic or a series of images in the same style.
5. Get Consistent Characters with Character References (--cref)
This is the “holy grail” for storytellers and brand builders. It solves the biggest problem in AI art: getting the same character to appear in different scenes.
What it does: The --cref parameter allows you to use an image of a person or character as a “character reference.” Midjourney will try to replicate the character’s key features (face, hair, clothing) in your new prompt.
How to use it:
- Generate an image of your character. Upscale your favorite one to get a clean, clear shot.
- Copy the URL of that upscaled image.
- Write your new prompt with the character in a different scene.
- Add the parameter:
--cref https://imagedescriber.online/blog/article/ai-character-photo-describer
- Prompt 1 (Create Character):
/imagine prompt: a female cyberpunk detective, close up portrait, purple hair, neon city - (…you upscale your favorite image and get its URL…)
- Prompt 2 (Use Character):
/imagine prompt: a female cyberpunk detective walking down a dark alley --cref https://.../your-character-image.jpg - What this does: It generates a new scene of your character walking, but it will try to match the face, hair, and style from your first image. It’s not perfect, but it’s a massive leap forward for creating consistent characters in a series.
Conclusion
Mastering these 5 advanced techniques—Multi-Prompting, Parameters, Permutations, Style References, and Character References—elevates you from a simple prompter to an AI art director.
You are no longer just rolling the dice; you are giving Midjourney specific, nuanced instructions to get the exact image you have in your head. The key is to start experimenting. Combine these techniques, test different parameters, and see how you can push the boundaries of your own creativity.
What are your favorite “secret” Midjourney techniques? Share your best tips or prompts in the comments below!
Leave a Reply