- Imagine Weekly – The only Midjourney Newsletter you need
- Posts
- Imagine Weekly – Issue 2 [June 13th]
Imagine Weekly – Issue 2 [June 13th]
Welcome 👋 Great to have you back 🤖🎉
Hi, and welcome back to this week's issue of Imagine Weekly. I'm thrilled to see our subscriber list growing. I'm eager to deliver great content in every issue. If there's anything you're missing or if you'd like to see more of a specific category, please feel free to reach out. Your input can make Imagine Weekly even better 💯🚀
A new Feature has arrived
Ok so this week well kick this issue of with a great new feature: MidJourney's New Personalization Feature: Tailor Your Image Generations.
MidJourney has just rolled out an exciting new personalization feature that promises to revolutionize your image generation experience. This innovative update allows the bot to take into account your unique preferences, ensuring that the images it creates are more aligned with your personal style. Here’s a quick rundown of what this feature entails and how you can make the most of it.
What is the Personalization Feature?
Activation: You can activate personalization by adding -P to your prompts or by enabling it in your settings. This tells the bot to use your personal style code when generating images.
Learning Your Preferences: The bot learns your preferences based on the images you like and rank on the MidJourney website. By selecting around 200 images, the bot can start to understand your style. The more you rank, the better it gets—some users have ranked thousands of images to fine-tune their personalization.
Strength Control: You can control the influence of personalization using a stylized value (-S), which ranges from 0 to 1,000. The default value is 100, but you can adjust it to see more or less of your personal style in the generated images.
Chaos Value: Experimenting with chaos values can yield different variations in your images. Higher chaos values can produce more unexpected and creative results.
Aspect Ratios: Be aware that the aspect ratio of your image can affect the personalization results. Different ratios might emphasize different elements of your style.
Short vs. Long Prompts: The personalization feature works well with both short and long prompts. Even with detailed prompts, your personal style will shine through in the generated images.
Style Raw Compatibility: Your personal codes also work with the style raw feature, allowing for even more nuanced and tailored image generations.
How to Get Started
Enable Personalization: Go to your settings on the MidJourney website and turn on the personalization feature.
Rank Images: Start ranking images on the explore page to help the bot learn your preferences.
Use Your Code: Add -P to your prompts to activate your personal style code.
Adjust Strength: Use the stylized value (-S) to control how much of your personal style is applied.
Experiment: Play around with chaos values and different aspect ratios to see how they affect your results.
Example
Imagine you prompt for "an aerial view of a pirate ship festival cruise." Normally, MidJourney would generate a standard image. However, by adding -P to your prompt, the bot will personalize the image based on your preferences. You can further adjust the stylized value to see more or less of your personal style in the final image.
Final Thoughts
This personalization feature is off by default, so you don’t have to use it if you prefer the standard image generations. However, if you want to create images that truly reflect your unique style, this feature is a game-changer. Trust the process, experiment with different settings, and enjoy the creative possibilities that MidJourney’s personalization feature offers.
Happy creating!
If you want to get a quick video intro check out Future Tech Pilot great video: 📺 https://www.youtube.com/watch?v=kC2SyWMZ_Rg
Office Hours Recap
You can skip the TL;DR; and read the full recap down below.
TL;DR;
📢Recent Releases: New Niji version with anime aesthetics and a personalization feature.
🔜Upcoming Releases: Enhancements in blending and personalization, with significant improvements in Versions 6.1/6.5 and 7.
🌐Website Changes: Simplifying user experience and adding more community features.
📊Surveys: Demographic and psychometric surveys to tailor the platform to user preferences.
📚 Magazine/Book: Hardcover "Super Edition" book celebrating the first year, coming soon.
Extended Recap
📢 Recent Releases
New Niji Version: Features anime/Eastern aesthetics, supports Kanji and other Japanese/Chinese characters. Users are enjoying it.
Personalization Feature: Customizes the model to user preferences based on image ratings and comparisons. Currently off by default but may be on by default in V7. Developed over a year, it enhances image quality by aligning with user tastes.
🔜 Upcoming Releases
Style and Personalization Codes: Combining these for easier blending, expected next week.
Rapid Personalization Updates: More frequent updates based on user actions, expected next week.
Version 6.1 or 6.5: Final training run starting, will take 2-4 weeks. Testing V7 features in V6, with improvements in image quality, prompt understanding, and aesthetics. Potential upscaler update.
Version 7: Not in final training yet, will take at least a month once started. Debating conservative vs aggressive approach. Will include improvements in image, prompt, knowledge, and aesthetics, with potential speed and controllability enhancements.
🌐 Website Changes
Simplification for New Users: Many small changes aimed at making the site easier for new users who find Discord confusing.
Slow Rollout: Changes are being rolled out slowly with small cohorts to track metrics versus Discord.
Community Features: Plans to add more community features, resources, and explorer functionality.
MVP for Public Access: Moving towards a Minimum Viable Product for open public access between V6 and V7.
Idea Submissions: Another round of idea submissions and voting will occur after V7.
📊 Demographic and Psychometric Surveys
Demographic Survey: Coming this week, optional but offers Fast Hours for participation. Aims to study the connection between aesthetics and demographics, potentially allowing demographic parameters on prompts. Example: Optimizing for colorblind users.
Psychometric Survey: Coming later, more focused on psychological aspects to understand the origins of aesthetics.
Data Security: Survey results can be deleted by users, and data will be kept secure like billing info.
Articles
Yes, you read that right, there's a Miss AI pageant. The inaugural "Miss AI" pageant, where AI-generated models compete using photorealistic images on social media platforms like Instagram, debuts this month. These virtual contestants, a product of various AI technologies, post updates and interact with audiences online. Despite its digital nature, the competition offers tangible rewards, such as a $5,000 cash prize and mentorship opportunities. However, the fact that all 10 finalists adhere to conventional beauty standards has prompted conversations about diversity in AI content. The event underscores the marketing potential of AI, its transformative impact on beauty pageants, and raises concerns about inclusivity in AI-generated personas.
@Yana G.Y. recently encountered a series of bizarre and distorted image generations, including squint-eyed humans, disproportionate body parts, and even three-armed figures, despite using detailed prompts. After investigating, the user discovered that these anomalies were linked to using non-default aspect ratios. By switching back to the default square aspect ratio, the user was able to resolve the issues and achieve more accurate results. This experience highlights the importance of aspect ratio settings in AI-generated art and offers a useful tip for other MidJourney users facing similar problems.
Inspiration, prompts & Srefs
For today's issue, I've gathered some nice references for you to save. All images were generated by me, using either full or partial prompts from the original authors.
Tokyo Night / sref 1193720432
I discovered this effect myself while experimenting with this great feature. It works perfectly for night scenes. I gave it a shot while reminiscing about my time in Japan.
eating out at shibuya station at night
Summer Landscapes / sref 700561985 by @design_maverck
LA Mountains wir the Hollywood sign, palm trees and the blue sky --ar 16:9 --style raw --sref 700561985
See you next week 👋🤖
I hope you enjoyed this issue of Imagine Weekly. If you found it helpful, we'd be thrilled if you could share it with your colleagues and friends. Please feel free to reach out with any suggestions or feedback.