Midjourney v6 News: SREF v2 Released

Thaeyne
15 Mar 202405:38

TLDRThe latest update on Midjourney's style reference (SREF) has been released, with version 2 offering improved precision in understanding and applying style to generated images. David Holtz explains that the new version is better at excluding non-stylistic elements. The functionality remains largely the same, with users able to apply style references using the 'sref' command and adjust the style weight from 0 to 1,000. The old version is no longer available. The video provides a detailed exploration of how the style reference works, using various prompts and comparing the results with different style weights. The presenter tests the feature by generating images with prompts like 'fantasy' and 'eldrich', and adjusting the style weight to observe the impact on the generated content. The video concludes with a discussion on the challenges of understanding the numerous combinations of value scales and encourages viewers to share their thoughts on the new version and their usage of different scales.

Takeaways

  • 🎉 SREF v2 has been released, offering a more precise understanding of style and better control over non-style elements in generated images.
  • 📈 The new version of SREF is said to be more accurate, as mentioned by David Holtz, and it's designed to prevent non-stylistic elements from leaking into the images.
  • 🔄 The functionality of Style Reference remains the same, with the ability to use it with 'sref' and 'ssw' for style weight control from 0 to 1000.
  • 🚫 The old version of Style Reference is no longer available.
  • 🌊 In an example, the word 'fantasy' generated atmospheric and painterly images, while adding 'eldrich' resulted in darker colors and more tentacles.
  • 🔄 By using the entire grid as a style reference for 'eldrich', the new version two accurately captured the desired style with appropriate colors and fewer unintended elements.
  • 🐱 The 'eldrich drawing of a cat' prompt combined with the style reference resulted in a muted color palette and no tentacles, indicating the style was applied without overtaking the original prompt.
  • 🔄 Experimenting with different style weights showed that at lower weights, the style was subtly integrated, while higher weights led to more extreme transformations.
  • 📊 At style weight 200, the images began to lose their original form, and by 500, they were fully transformed into tentacle monsters, demonstrating the power of the style weight setting.
  • 🤔 The multitude of value scales and their combinations make it challenging to predict and understand their behavior, suggesting that experimentation is key.
  • 📌 The speaker recommends using either default or moderate values for the best results, rather than extreme ends of the scale.
  • ✅ The audience is encouraged to share their thoughts on the new version of Style Reference and whether they noticed a difference in its performance.

Q & A

  • What is the latest update regarding Midjourney's style reference?

    -The latest update is that Midjourney's style reference has been updated to version 2, which is reported to be more precise in understanding style and better at not leaking non-style elements into the generated images.

  • Who is credited with the announcement of the new version of the style reference?

    -David Holtz is credited with the announcement of the new version of the style reference.

  • How does the new version of the style reference work in comparison to the old version?

    -The new version of the style reference works similarly to the old version, but with improvements in precision and less leakage of non-style elements. The old version doesn't seem to be available anymore.

  • What are the available commands to use with the new style reference?

    -The available commands with the new style reference are 'sref' and 'ssw', which are used for style reference and style weight, respectively.

  • What is the range for the style weight in the new version of the style reference?

    -The style weight in the new version of the style reference can range from 0 to 1,000.

  • How does setting a seed number affect the image generation process?

    -Setting a seed number ensures that all generated images are based on the same seed number, allowing for consistent results and making it easier to observe the effects of changes made to the prompt or style reference.

  • What was the result of adding the word 'eldrich' to the word 'fantasy' as a prompt?

    -Adding the word 'eldrich' to the word 'fantasy' as a prompt resulted in images with darker colors and slightly more tentacles, indicating a more ominous and atmospheric style.

  • How does the style reference affect the image generated with the prompt 'drawing of a cat'?

    -When the style reference is applied to the prompt 'drawing of a cat', the resulting image maintains some characteristics of a drawing but introduces muted colors and no tentacles, indicating a successful application of the 'eldrich' style.

  • What happens when the style weight is increased from 0 to 1,000 with the 'eldrich drawing of a cat' prompt?

    -As the style weight increases from 0 to 1,000, the images progress from a slight change with only a tiny difference to extreme transformations, eventually resulting in tentacle monsters with no trace of the original cat or drawing.

  • How does the new version of the style reference handle different value scales?

    -The new version of the style reference handles different value scales by offering a wide range of adjustments, making it challenging to understand how each scale behaves individually and in combination with others. It is suggested that the best results are achieved with moderate values rather than extreme ends of the scale.

  • What is the speaker's recommendation for using the different scales in the style reference?

    -The speaker recommends using the default value or moderate values for the different scales in the style reference, as extreme values may not yield the best results.

  • How can viewers engage with the content and the speaker?

    -null

Outlines

00:00

🎨 Mid Journey Style Reference Update

The video discusses an update to the Mid Journey style reference, which has been upgraded to version 2. The update is said to improve the precision of style understanding and reduce the inclusion of non-style elements in image generation. The style reference continues to function similarly to its previous iteration, with adjustments made through the '--sref' and '--ssw' parameters. The creator demonstrates the functionality of the style reference through various examples, including the generation of images with a fantasy theme and the application of an 'eldrich' style reference to alter the output. The video also explores how different style weights affect the final image, showing a progression from subtle changes at lower weights to more dramatic transformations at higher weights. The creator concludes by inviting viewers to share their thoughts on the new version of the style reference and their preferences for using different scales.

05:02

📢 Viewer Engagement and Subscription Call

The video concludes with a call to action, encouraging viewers to subscribe to the channel for more content like the one they just watched. The creator also expresses appreciation for likes and views, and thanks the audience for their time. The video ends with a short musical interlude, indicating the conclusion of the content.

Mindmap

Keywords

Midjourney v6

Midjourney v6 refers to the sixth version of a software or tool, presumably related to image generation or design. It's the main subject of the video, as the update and its features are being discussed. The video provides information about the improvements and changes in this version.

SREF v2

SREF v2 stands for 'Style Reference version 2', which is an updated feature of the Midjourney v6 software. It is designed to be more precise in understanding and applying styles to generated images without introducing unrelated elements. The video demonstrates how this feature works and its impact on image generation.

Style Weight

Style Weight is a parameter in the Midjourney v6 software that allows users to control the intensity of the style being applied to an image. It ranges from 0 to 1,000, with different values producing varying degrees of stylistic influence. The script discusses how adjusting this weight can lead to different outcomes in image generation.

Image Generation

Image Generation is the process of creating visual content using the Midjourney v6 software. It is central to the video's theme as the host demonstrates how different prompts and style references affect the generated images. The video shows examples of image generation with various settings and parameters.

Prompt

A Prompt is an input or instruction given to the Midjourney v6 software to guide the image generation process. The video script describes how different prompts, such as 'fantasy' or 'eldrich', can lead to different styles and elements in the generated images.

Seed Number

The Seed Number is a value set by the user that determines the randomness in the image generation process. In the context of the video, the host uses the same seed number (777) for consistency, allowing changes in the prompt or style reference to be the sole factors in variations of the generated images.

Eldrich

Eldrich, in the context of the video, is a stylistic term used as a prompt in the image generation process. It is associated with darker colors and elements like tentacles in the generated images. The video explores how the term 'eldrich' influences the style of the generated images when used in conjunction with other prompts.

Fantasy

Fantasy, as a prompt, is used to generate images with a fantastical theme. The video compares the results of using 'fantasy' alone to using it in combination with 'eldrich' to demonstrate how the style reference can modify the generated images to match a specific aesthetic.

Tentacles

Tentacles are a recurring element in the generated images that become more prominent as the 'eldrich' style is applied. They serve as a visual example of how the software interprets and incorporates stylistic elements based on the prompts and style references used.

Styliz Value

Styliz Value, while not explicitly detailed in the script, can be inferred as a related parameter to Style Weight that affects the stylistic output of the image generation process. It is mentioned in comparison to the Style Weight, suggesting it's another way to control the style application in image generation.

Combination of Scales

The Combination of Scales refers to the various settings and parameters that can be adjusted in the Midjourney v6 software to achieve different results in image generation. The video emphasizes the complexity and near impossibility of testing every possible combination due to the vast number of options available.

Highlights

Midjourney style reference (SREF) has been updated to version 2.

The new version is said to be more precise in understanding style and better at excluding non-stylistic elements from images.

David Holtz mentioned that the functionality of style reference remains largely the same, with improvements on the backend.

The old version of SREF is no longer available.

Users can still use the style reference with 'sref' and have 'ssw' available for style weight adjustments.

A demonstration of image generation using the prompt 'elrich' resulted in Krakens in a dark green or blue setting.

Comparing the results of the 'fantasy' prompt to 'eldrich fantasy' showed darker colors and more tentacles.

Using the entire grid as a style reference for 'eldrich' produced images with accurate colors and fewer unwanted elements.

The older version might have included unintended elements like tentacles even when not prompted.

Adding 'eldrich' to the prompt while keeping the style reference constant resulted in stylistically similar images to the reference.

The 'fantasy' and 'eldrich' styles are noted to be quite similar stylistically.

An experiment with 'eldrich drawing of a cat' produced cat monsters with tentacles.

The style weight's impact on the image generation was tested at various breakpoints, from 0 to 1,000.

At a style weight of 10, the images showed small changes, making the cats look slightly scarier.

At style weight 50, the images began to show more tentacles, indicating a shift from cute to monstrous.

At the default style weight of 100, the images were monstrous but still recognizable as cats.

Extreme style weights like 200 and 500 led to images dominated by tentacles, losing the original cat or drawing characteristics.

At style weight 1,000, the images were entirely tentacle monsters, indicating the importance of choosing the right style weight.

The complexity of different value scales and their combinations makes it challenging to predict and understand their behavior.

The video suggests that the best results are achieved with the default value or moderate adjustments rather than extreme ends of the scale.

The presenter invites viewers to share their thoughts on the new style reference version and their usage of different scales in the comments.