Midjourney v6.1 vs v6.0: What You NEED to Know | SREF, Beauty Bias, New --p Codes | Midjourney Guide

Woollyfern
8 Aug 202406:18

TLDRMidjourney's v6.1 model offers improved image quality with enhanced textures and realism, especially in eyes and skin. It introduces a Q2 mode for more detailed results, albeit with slightly reduced coherence. The new upscaler is comparable to Topaz Labs' Gigapixel, with each showing strengths in different areas. While v6.1 exhibits a conventional beauty bias, making characters appear smoother, Midjourney is aware and plans to address this in v7. The model also supports character references and style references (SR codes) with some stylistic differences. Personalization has been enhanced, with improved text rendering and a new model personalization system that adapts based on user preferences.

Takeaways

  • 🆕 The new v6.1 model is now the default for all users, but you can change it in your settings.
  • 🖼️ Image quality in v6.1 has improved with reduced pixel artifacts and more realistic eyes and skin textures.
  • 🔍 There's a new Q2 mode in v6.1 that can enhance texture and details at the cost of slight image coherence reduction.
  • 🆚 Midjourney's 2X upscaler is compared with Topaz Labs' Gigapixel, with each having their strengths in different areas.
  • 💆‍♀️ v6.1 has a more conventional beauty bias, resulting in smoother skin and fewer imperfections, which is being addressed for v7.
  • 🎭 You can use previous character references with v6.1, but there might be a slight increase in conventional beauty bias.
  • 🎨 Style references (SR codes) should work similarly in both v6.0 and v6.1, but v6.1 might produce less detailed images.
  • 📝 Text rendering accuracy has improved in v6.1, though more user feedback is needed to confirm.
  • 🧩 The tile parameter for creating seamless patterns works better in v6.1 and integrates with the new upscaler.
  • 🤖 The model personalization system in v6.1 is more accurate and nuanced, adapting to your aesthetic preferences based on image rankings.
  • 🔄 You can use both old and new personalization codes in v6.0 and v6.1, but mixing codes from different systems is not allowed.

Q & A

  • What is the main difference between Midjourney v6.1 and v6.0 models?

    -The main differences include improved image quality with reduced pixel artifacts and enhanced textures in v6.1, as well as a more conventional beauty bias in the images it produces.

  • How does the image quality differ between Midjourney v6.1 and v6.0?

    -v6.1 brings a boost to image quality with better clarity in eyes and skin, improved coherence and accuracy in bodies and hands, and reduced pixel artifacts.

  • What is the purpose of the '--Q2' parameter in Midjourney v6.1?

    -The '--Q2' parameter in v6.1 can be added to prompts to sometimes get more texture and details in the results, albeit at the cost of slightly reduced image coherence.

  • How does Midjourney v6.1 compare to Topaz Labs' Gigapixel in terms of upscaling?

    -Both Midjourney's upscaler and Topaz Labs' Gigapixel do a decent job, but the video suggests that Topaz may handle skin texture and eye clarity better in some cases.

  • What is the conventional beauty bias mentioned in the v6.1 model?

    -The conventional beauty bias in v6.1 refers to the model's tendency to produce images with smoother skin and fewer imperfections, which was an unintended side effect of incorporating user-generated data to boost image quality.

  • How can users switch back to the v6.0 model if they prefer it over v6.1?

    -Users can set a new default model in their settings if they prefer the v6.0 model over the v6.1 model that is set as default.

  • Can character references from previous versions be used in Midjourney v6.1?

    -Yes, users can use any of their previous character references with v6.1, and it handles character references similarly to v6.0.

  • How does the style rendering differ between v6.0 and v6.1 when using SR codes?

    -While the style should remain the same between v6.0 and v6.1, v6.1 images may appear slightly more simplistic and less detailed compared to v6.0.

  • What improvements are expected in the text rendering accuracy in Midjourney v6.1?

    -v6.1 is expected to have improved text rendering accuracy, handling longer phrases with fewer mistakes and more complicated phrases better.

  • How does the new model personalization system in v6.1 work?

    -The new model personalization system in v6.1 is more accurate and nuanced, interpreting visual aesthetic preferences based on image rankings. Users can see all their model personalization codes by typing /list personalized codes in Discord.

  • What is the benefit of the improved 'tile' parameter in v6.1?

    -The 'tile' parameter in v6.1 works better and maintains seamless patterns even when images are upscaled, allowing for higher resolution tiles without breaking the pattern.

Outlines

00:00

🖼️ Mid Journey v6.1 Model Overview and Comparison

This paragraph introduces Mid Journey's v6.1 model, comparing it with the previous v6.0 model. The v6.1 model offers improved image quality with reduced pixel artifacts and enhanced textures, particularly in eyes and skin clarity. It also includes a Q2 mode for additional texture details at the cost of slight coherence reduction. The paragraph discusses the new 2X upscaler and compares it with Topaz Labs' Gigapixel, noting that the choice between them depends on the specific image and purpose. The v6.1 model has a more pronounced conventional beauty bias, which was acknowledged by Mid Journey and is set to be corrected in the upcoming v7 model. The paragraph concludes with a discussion on character references and how they remain consistent between v6.0 and v6.1, despite minor differences in conventional beauty bias.

05:00

🎨 Exploring Mid Journey v6.1's Style References and Personalization

The second paragraph delves into the use of style references with the v6.1 model, noting that while the style remains consistent between v6.0 and v6.1, the new version tends to produce simpler and less detailed images. It also addresses improvements in text rendering accuracy in v6.1 and the seamless pattern creation with the 'tile' parameter. The paragraph highlights the new model personalization system, which is expected to be more accurate and nuanced based on user preferences. The speaker encourages viewers to test both old and new personalization codes to find the best fit for their needs. The paragraph ends with a personal preference for the v6.0 aesthetic due to its creativity, detail, and diversity, and anticipation for future improvements in v7. It also mentions upcoming releases of new srf codes and a monthly prompt collection for Patreon members.

Mindmap

Keywords

💡Midjourney v6.1

Midjourney v6.1 refers to the latest version of the AI image generation model by Midjourney, a company specializing in AI-generated content. This version is compared to the previous v6.0 in the video, highlighting improvements in image quality, reduced pixel artifacts, and enhanced textures. The video discusses how v6.1 has become the default model for all users, and how it handles character references and style references with examples from the script.

💡Image Quality

Image quality is a critical aspect of AI-generated imagery, focusing on the clarity, realism, and coherence of the generated images. The video script mentions that v6.1 brings a boost to image quality with better clarity in eyes and skin, and improved coherence in bodies and hands, which are significant for users looking for more realistic and detailed outputs.

💡Pixel Artifacts

Pixel artifacts are visual imperfections that can occur in digital images, appearing as blockiness, blurriness, or other distortions. The script notes that Midjourney v6.1 reduces these artifacts, leading to cleaner and more polished images, which is a significant advancement for users who value high-quality visual outputs.

💡Upscaler

An upscaler is a tool or feature that increases the resolution of an image without losing quality. The video discusses Midjourney's new 2X upscaler and compares it to Topaz Labs' Gigapixel, indicating that v6.1 offers improved upscaling capabilities. This is important for users who need to enlarge their images without compromising on detail and clarity.

💡Conventional Beauty Bias

Conventional Beauty Bias refers to the倾向性 of AI models to generate images that adhere to widely accepted beauty standards, often leading to a lack of diversity in appearance. The script points out that v6.1 has a more pronounced conventional beauty bias compared to v6.0, resulting in smoother skin and fewer imperfections, which is a concern for those seeking more varied and unique character representations.

💡Sr Codes

Sr codes, or style reference codes, are specific parameters used in Midjourney to generate images in a particular artistic style. The video explores how these codes perform in v6.1, noting that while the style remains consistent between versions, v6.1 tends to produce simpler and less detailed images, which could be a consideration for users looking for specific levels of detail and complexity.

💡Model Personalization

Model personalization in Midjourney allows users to train the AI to better understand and generate images according to their aesthetic preferences. The video explains that v6.1's model personalization system is more accurate and nuanced, adapting to user feedback through image rankings. This feature is significant for users who want a more tailored and personalized image generation experience.

💡Text Rendering Accuracy

Text rendering accuracy pertains to the AI's ability to correctly and clearly generate text within images. The script suggests that v6.1 has improvements in this area, which is crucial for users who incorporate text into their image generation, requiring high precision and readability.

💡Tile Parameters

Tile parameters are used in Midjourney to create seamless patterns that can be tiled without visible seams. The video notes that v6.1 resolves issues with upscaling breaking the seamless pattern, allowing for higher resolution tiling without loss of pattern integrity. This is beneficial for users who work with patterns and textures in their image generation.

💡Midjourney Office Hours

Midjourney Office Hours refer to a time when the company's team is available to address user questions and concerns. The script mentions that the conventional beauty bias in v6.1 was discussed during such an event, indicating that the company is aware of the issue and plans to address it in future updates, which is important for users who value transparency and ongoing development.

Highlights

Midjourney's new v6.1 model is now the default for all users, but can be changed in settings.

v6.1 enhances image quality with reduced pixel artifacts and improved textures.

Eyes and skin in v6.1 look more realistic and have improved clarity.

v6.1 offers better coherence and accuracy for body and hand depictions.

The --Q2 mode in v6.1 provides more texture and details at the cost of slightly reduced image coherence.

New 2X upscalers in v6.1 are compared favorably with Topaz Labs' Gigapixel.

v6.1 has a more pronounced conventional beauty bias, with smoother skin and fewer imperfections.

User-generated data was used to improve v6.1's image quality, inadvertently increasing the beauty bias.

v6.0 may be preferred for creating images of average or unique-looking people due to v6.1's beauty bias.

Character references work similarly in both v6.0 and v6.1, with minor differences.

SR codes maintain the same style in v6.1 but with less detail and complexity.

Text rendering accuracy is reportedly improved in v6.1.

The tile parameter in v6.1 maintains seamless patterns even after upscaling.

The new model personalization system in v6.1 is more accurate and nuanced.

Personalization codes change as you rank more images in v6.1.

Old and new personalization codes can be used in both v6.0 and v6.1, but not combined across versions.

v6.1 offers improved quality and clarity, but some prefer v6.0 for its creativity and diversity.

The video creator looks forward to the upcoming V7 release, which may address v6.1's limitations.

Casual Browsing

Stable diffusion VS Midjourney: All you need to know

2024-05-22 14:50:02

Midjourney Copyright Law: What AI Artists Need to Know

2024-05-19 10:00:01

6 Essential Midjourney Parameters You Need To Know

2024-05-23 08:30:01

Midjourney On The Web - All You Need To Know | Midjourney Alpha

2024-06-22 21:20:00

The Midjourney Showcase in-depth: All you need to know.

2024-05-18 23:55:02