According to previous updates, Midjourney has just launched a new "Style Reference" that brings additional aesthetics with a new “Versioning” feature. “Style References” gives you the ability to reference the style of another image or to keep a consistent style across many images.
Today, I will walk you through the new aesthetics and share my 3 Advanced Skills to leverage “Style Reference”!
Welcome aboard my aircraft. Our exciting journey is about to begin! 🚀
Today’s content will be broken down into four sections:
Part One: What is “Style References”
Part Two: How to use “Style References”
Part Three: New Changes — 4 Versioning Feature with --sv
Part Four: Three Advanced Tips to Level Up
🎨 Part One: What is “Style Reference”
“Style References” give you the ability to reference the style of another image or to keep a consistent style across many images.
The “--sref <URL>" parameter allows us to create a new image using the visual style of an existing image.
Functionality: Work similar to image prompts where you give a URL to one or more images that ‘describe’ the consistent style you want to work over.
Usage: Type --sref URL after your prompt with a URL to an image.
Format: /imagine text prompt --sref stylePrompt1 stylePrompt2
Compatibility: Works for both MJ V6 and Niji V6
🎨 Part Two: How to use “Style Reference”
Step 1: Pick Your Reference Image
Pick an image you generated with Midjourney or pick one online.
Step 2: Upload the Image
Identify one or more style reference image URLs that end in a filename such as jpg, gif, webp, or png.
Click the [ + ] button in your chat bar and upload an image
Right-click the center of the image and select ‘Copy Link’ or ‘Copy Media Link’ .
Step 3: Type Your Prompt
Type/imagineas you usually do. Then, input your prompt to describe the image you want to generate. Type --sref after your text prompt, followed by URLs
/imagine prompt: Text Prompt --sref url
Surreal concept art of a submarine --sref url
Magic of using “Style Reference” 💫
The magic of using a “style reference” image (--sref) is that Midjourney will strive to pick up only the aesthetic qualities of a reference image instead of its semantic content as well.In this sense, we can think of --sref like a Text Expander Type Shortcut.
💫 Part Three: New Change — “Versioning Feature” with --sv
The latest version of the style reference algorithm (now default) is
More precise about understanding style
Improved weight parameter --sw , more meaningful with a bigger range.
Come with 4 versions with --sv
Four “Versions“ with --sv🔥
If you want to explore outside the default, you can now use
--sref <URL> with --sv N (sref version) where N is 1, 2, 3, or 4:
--sv 1 The original style reference
--sv 2 The one that was the default
--sv 3 A new version of sv1 with better performance and more 'vibey' than sv2
--sv 4 A new version of sv2 with improved performance ( default )
Here are 4 Different Versions generated using the hot air balloon as a reference image.
Text Prompt --sref <URL> --sv {1,2,3,4}
Personally, I prefer sv2. I suggest having more trial and experiment and explore the one that suits your needs the most. 😊
📚 Part Four: Three Advanced Tips to Create the Image in Your Mind
⭐️ Tip 1: Solving Undesirable Details
Note that “Style Reference” applies ONLY the aesthetics and NOT the content.
In some cases, Midjourney will also pick up some of the content that the new prompt didn’t specify.
Three methods to explore:
If you want to get away from the style reference image accidentally adding undesirable details to your canvas, there are three methods you can explore.
📌 Method 1: Adjust the prompt.
👉🏻 Too many unwanted semantic details?
✅ Write a prompt that replaces the undesirable details with ones that you do find desirable.
👉🏻 Not enough of the style?
✅Modify your prompt to lean in to the style you’re chasing.
✅ If you want textured, impasto, expressive oil painting and it is not showing up strongly enough, add the words to the prompt in solidarity with your sref images.
For instance, you can find that the images generated above all contain clouds but you are looking for a scene in the sea
New Prompt:
Rococo art painting of a submarine in the sea with an underwater city below it --sref url
By adjusting the prompt, you have the flexibility to eliminate unwanted semantic details and infuse your preferred style seamlessly.
You can also try adjusting the image weight
📌 Method 2: Lower Image Weight (--sw)
👉🏻 Too many semantic details? ✅ Lower your --sw from the default of 100 to something less dominating, like --sw 20.
Use small weights to emphasize or de-emphasize elements on the canvas you find undesirable.
📌 Method 3: Raise Image Weight (--sw)
👉🏻 Not enough of the style? ✅ Raise your --sw from the default of 100 to something more dominating, like --sw 800 (max is 1000).
Here are examples with different image weights:
In addition to adjusting text prompts and weight, we can also use “Image Prompt” and “Character Reference” to influence the details in the image
⭐️ Tip 2: Enhance Your Creativity with “Image Prompts”
While the Image Prompt feature is excellent for shaping the composition of your image, you can elevate it further by incorporating a specific style using “Style Reference.”
This is great for replicating an image with a new style.
📌 What is Image Prompt
You can use images as part of a prompt to influence a Job’ s composition. Images prompts can be used alone or with text prompts.
Image Prompt + Style Reference
Use: To use an image prompt, it must go before --sref
Format: Text Prompt ImagePrompt --sref urlA
A surreal Hot Air Balloon
/imagine Surreal concept art of a hot air balloon ImagePrompt --sref URL
Here is the resulting image.
Breakdown
Image Prompt: Replicate the composition of the hot air balloon
Style Reference: Replicate the art style and aesthetics
⭐️ Tip 3: Combine the Power of “Style Reference” & “Character Reference”
You can use more a reference image to influence the style of your character. This is great for generating a set of collections or graphic novels.
📌 What is Character Reference
“Character Reference” allows us to create a character based on the image(s) of an existing character provided via URL with --cref
Character Reference + Style Reference
Use: Type --cref & --sref after your text prompt, followed by URLs to more than one reference image like this
Format: Text Prompt --cref urlA --sref urlB
A guy reading
A handsome guy reading in the library --cref urlA --sref urlB
Here is the resulting image
Breakdown
Character Reference: Create a consistent character
Style Reference: Replicate the art style and aesthetics
This skill is great when you want to prompt multiple characters with a consistent style.
Comments