Midjourney's NEW Secret Weapon - Style Reference

Making AI Magic
7 Mar 202413:16

TLDRMidjourney introduces Style Reference (sref), a tool for version six and Niji version six, allowing users to replicate preferred styles by referencing images. This feature aids those who struggle to describe styles and facilitates style combination. Unlike image prompts, sref focuses on capturing the aesthetic rather than composition. Users can combine sref with other parameters like stylize and style raw for varied outcomes. The tool is particularly useful for creating consistent styles across a series, such as graphic novels. To use sref, an image URL is required, which can be inputted in Discord or on Midjourney's website. The innovative feature enables artists to direct the AI towards a desired aesthetic, streamlining the creative process.

Takeaways

  • 🎨 Midjourney introduced a new tool called Style Reference (sref) in February 2024 for version six and Niji version six.
  • 🖼️ Style Reference allows users to replicate a preferred style by using an image as a reference, simplifying the process for those who struggle to describe the style they want.
  • 🔄 The sref feature is distinct from image prompts, focusing more on capturing the style and look of the reference image rather than composition.
  • 🌟 Style Raw and Stylize parameters are popular for version six, with Style Raw reducing Midjourney's style preferences and Stylize adjusting the level of Midjourney's beauty enhancements.
  • 🔧 Version five offers the Style Tuner for selecting favorite styles based on image choices, and Niji 5 has anime-style options like scenic, cute, and expressive.
  • 📚 The goal of sref is to create consistent styles, which is useful for projects like graphic novels that require a uniform aesthetic throughout.
  • 🔗 To use sref, users need an image URL, which can be obtained by uploading an image or using an online image and copying its URL.
  • 🤖 Both Discord and the Midjourney website support the use of sref, with each platform having its own method for applying the style reference to prompts.
  • 🎯 The Style Weight (SW) parameter can be added to further emphasize the style reference, with a scale from 0 to 1000, allowing for greater control over the aesthetic outcome.
  • 🧩 Multiple style references can be combined in a single prompt, with the option to assign different weights to each reference to control their influence on the final image.

Q & A

  • What is Midjourney's new tool called?

    -Midjourney's new tool is called Style Reference, or sref.

  • When was the Style Reference tool released?

    -The Style Reference tool was released in February 2024.

  • Which versions of Midjourney support the Style Reference feature?

    -The Style Reference feature only works on Midjourney version six and Niji version six.

  • What is the main purpose of using Style Reference?

    -The main purpose of Style Reference is to help users replicate a specific style they like by using an image as a reference, making it easier for users who find it difficult to describe the style they want.

  • How does Style Reference differ from image prompts?

    -While image prompts focus mainly on composition and layout, Style Reference captures the overall look and style of the image to a greater extent, helping to maintain consistency in style across multiple generated images.

  • What are some of the style parameters available in Midjourney version six?

    -In Midjourney version six, some of the style parameters include style raw, stylize, and the new style weight (SW) parameter that works alongside Style Reference (sref).

  • How can you use Style Reference in Discord?

    -In Discord, you can use Style Reference by uploading an image, copying its URL, and then typing your text prompt followed by '--sref' and pasting the image URL in the message bar.

  • What is the role of the SW (style weight) parameter?

    -The SW (style weight) parameter adjusts the influence of the style reference image on the generated image. It is on a scale from 0 to 1000, with higher values increasing the style influence.

  • Can you combine Style Reference with other parameters like stylize and style raw?

    -Yes, you can use Style Reference alongside other parameters such as stylize, weird, chaos, and even style raw for interesting and varied outcomes.

  • How can multiple Style Reference images be used at once?

    -Multiple Style Reference images can be used by separating each image URL with a space or by dragging multiple images into the web prompt bar, and using weights to specify the influence of each reference image.

  • How does the character reference parameter (cref) mentioned in the script complement Style Reference?

    -The character reference parameter (cref) is an upcoming tool that will focus primarily on recreating a character reference, which can complement Style Reference by helping to maintain consistency in character depiction across images.

Outlines

00:00

🎨 Introducing Style Reference Tool (sref)

This paragraph introduces Midjourney's new Style Reference tool (sref), which was released in February 2024 for version six and Niji version six. The tool allows users to replicate a specific style they like by simply pointing at an image and instructing the AI to 'make it look like that.' This feature is particularly useful for those who struggle to describe the style they want or enjoy combining different styles. The paragraph explains that sref differs from image prompts, which mainly focus on composition, and that sref aims to capture the overall look and style of an image. It also mentions that sref is not a replacement for image prompts but an additional feature that can be used together for more nuanced results. The paragraph concludes by noting that sref is part of Midjourney's broader effort to develop algorithms for creating consistent styles, such as for a graphic novel series.

05:01

📸 How to Use Style Reference (sref) Parameter

This paragraph provides a tutorial on how to use the sref parameter in Midjourney, both on the Discord platform and the Alpha website. It explains the process of uploading an image to use as a style reference and how to combine it with a text prompt. The paragraph clarifies the difference between a style reference image and an image prompt on the website, and how to use both simultaneously for a more tailored output. It also introduces the Style Weight (SW) parameter, which allows users to adjust the influence of the style reference image on the final output. The paragraph encourages experimentation with sref, including using multiple style references and adjusting their weights to achieve the desired aesthetic.

10:02

💬 Combining Text Prompts with Style References

The final paragraph emphasizes the necessity of including a text prompt when using the sref parameter, as it is not possible to use a style reference image without some accompanying text. It guides users on how to incorporate the sref parameter in their prompts, whether on Discord or the Alpha website, and highlights that sref is not a style transfer tool but rather a reference for Midjourney to create new images. The paragraph also discusses the possibility of blending multiple style references and adjusting their weights for varying degrees of influence. It concludes by encouraging users to share their experiences with the style reference parameter and to collaborate in creating amazing outputs.

Mindmap

Keywords

💡Midjourney

Midjourney is an AI-based tool that generates images based on user prompts. In the context of the video, it is highlighted as a platform that has introduced a new feature called Style Reference (sref), which allows users to replicate a preferred style in their generated images. This tool is particularly useful for those who may struggle to describe the style they want in words, offering a more intuitive way to achieve desired visual outcomes by referencing an image.

💡Style Reference (sref)

Style Reference, or sref, is a parameter introduced in Midjourney's version six and Niji version six. It enables users to direct the AI to generate images in a particular visual style by referencing an image. Unlike image prompts that focus on composition, sref captures the overall look and aesthetic of the referenced image, helping to maintain a consistent style across a series of prompts. This tool is particularly beneficial for projects requiring a consistent stylistic theme, such as graphic novels.

💡Style Weight (SW)

Style Weight, or SW, is a parameter that works in conjunction with the Style Reference feature in Midjourney. It allows users to adjust the influence of the style reference image on the generated image. The SW parameter is on a scale from zero to 1000, with higher values increasing the prominence of the style from the reference image. This provides users with greater control over how much of the style they want to see in their final image.

💡Image Prompts

Image prompts in the context of Midjourney refer to the use of existing images to guide the AI in creating new images based on composition, layout, and to some extent, color and style. While image prompts help with the overall structure of the image, they primarily focus on how elements are arranged rather than capturing the exact visual style. This is where the Style Reference feature complements image prompts by focusing more on the aesthetic aspect.

💡Style Raw

Style Raw is a parameter in Midjourney's version six that minimizes the AI's stylistic preferences and focuses more closely on the user's prompt. While this can result in images that are more aligned with the literal description provided, it may not necessarily enhance the visual appeal of the images. Users have the option to add beauty to their images through descriptive words in the prompt.

💡Stylize

The Stylize parameter in Midjourney controls the level of AI's artistic enhancement to the generated images. With a default value of 100, users can reduce the stylize value to zero if they want minimal AI intervention, or increase it up to 1000 for more pronounced Midjourney aesthetic touches. The balance between stylize values and prompt understanding versus aesthetic quality is a key consideration for users looking to generate images that meet their specific creative needs.

💡Version Six

Version Six refers to the latest iteration of Midjourney's AI platform as of February 2024. It is the default version that includes new features such as Style Reference (sref) and the ability to use Style Weight (SW). This version represents the current state of the AI's capabilities, offering users more tools and options to fine-tune their image generation process.

💡Niji

Niji is a version of Midjourney that is specifically tailored for creating anime-style images. It includes unique style parameters such as Style Tuner and different anime style options like scenic, cute, and expressive. Niji version six, like Midjourney version six, supports the use of Style Reference (sref) to help users achieve a desired anime aesthetic in their generated images.

💡Style Tuner

Style Tuner is a feature from Midjourney's version five that allows users to select their preferred styles based on image choices. By choosing different versions of an image, users receive a unique style code that can be used to replicate the chosen style. This tool provides a way to customize the visual output of the AI based on personal preferences.

💡Character Reference Parameter (cref)

Character Reference Parameter, or cref, is an upcoming feature mentioned in the video that will focus primarily on recreating a character reference in the generated images. While the Style Reference (sref) parameter captures the overall style of an image, the cref is expected to help maintain consistency in character depiction across different images, which is crucial for projects like graphic novels or comic series.

💡Consistent Styles

Consistent styles refer to the ability to maintain a uniform visual aesthetic throughout a series of images or across multiple prompts. This is particularly important for projects like graphic novels or visual storytelling where a cohesive style helps in creating a unified narrative experience. Midjourney's Style Reference feature aids in achieving consistent styles by allowing users to reference a specific look or aesthetic that they want to replicate across different images.

Highlights

Midjourney introduces Style Reference (sref) tool for version six and Niji version six.

Style Reference allows users to replicate a preferred style by simply pointing at an image.

sref is particularly useful for those who struggle to describe the style they want with words.

The new tool differs from image prompts, focusing more on capturing the style rather than composition.

As of February 2024, sref is only compatible with Midjourney's default version six.

Style parameters like 'style raw' and 'stylize' offer control over the AI's creative influence on images.

Version five introduced the Style Tuner, allowing users to select favorite styles based on image choices.

Niji 5 provided anime-style options like scenic, cute, expressive, and original.

sref aims to create consistent styles, beneficial for projects like graphic novels.

Using an image prompt focuses mainly on composition, while sref captures the overall style.

sref and image prompts can be used together, focusing on different aspects of the inspiration image.

To use sref, you need at least one image URL, which can be from an online source or uploaded into Midjourney.

Discord and the Alpha website offer different methods for applying the sref parameter.

An image can serve both as an image prompt and a style reference, with the website using a paperclip icon to distinguish sref images.

The Style Weight (SW) parameter allows users to adjust the influence of the style reference image.

sref can be combined with other parameters like stylize, weird, chaos, and style raw for varied outcomes.

Multiple style references can be blended, with weights assigned to each to indicate their significance.

The style reference tool is designed to quickly achieve a desired look without the need for detailed verbal descriptions.

The introduction of sref represents a significant advancement in AI's understanding and application of visual styles.