🚀 Midjourney Office Hours Highlights: June 19th, 2024

The latest Midjourney Office Hours brought a few exciting updates and future plans. Here’s a comprehensive summary of what was shared. If you prefer to just get the gist of it, jump down to the TL;DR; section.

Extended Recap

 🎨 Personalization Feature

Midjourney recently added a personalization feature that's been well-received by users. To personalize the model, you need to provide at least 200 ratings at midjourney.com/rank. Most users prefer the personalized model. The team is working on improving this feature and is considering an update this week, though it might affect compatibility with older versions.

The long-term plan is to make personalization the default for everyone. They also plan to introduce semi-personalization options, like profiles based on demographics (e.g., male in America) and a version for colorblind users. The team is also exploring how aesthetics vary by demographics and aims to show Midjourney outputs based on different groups (e.g., Japan, America, men, women).

Personalization is still being refined. The team is learning from users who don't like it, noting these users often have higher error rates and different tastes. They're working to make personalization more stable, safe, and high-quality.

Let’s see what they’ll come up with. I think that the personalization feature is quite nice and the results are great. However, currently I always create two versions of a prompt, one with and one without personalization to see where things are going.

🔄 Blending Multiple Codes

Next they talked about updating its personalization code system to allow for the blending of multiple codes. Although blending weights will not be included for now due to complexity, users are encouraged to personalize their models, share them with friends, and experiment with blending codes. This update should be right around the corner. It’s expected to launch in a few hours after final QA (Quality Assurance).

📸 Upcoming Model Releases

Now comes the juice part we’re all excited for: new model releases are on the horizon. Midjourney 6.5 is currently in training, with a focus on improving image quality and addressing edge cases. The upscaler quality is also being updated, and the release is likely in a few weeks due to the long training time.

Meanwhile, Midjourney Version 7 is in progress, with individual parts coming online and looking promising. Datasets are being prepared for final training runs. V7 is expected to be a dramatic improvement over V6, aiming to be smarter, have better image quality and aesthetics, and be faster and cheaper. There is also a possibility of creating a slower, more expensive, higher-quality version of V7 after the initial launch.

I'm really excited that it's getting smarter (Whatever that means). Sometimes, when you need something specific in an image, it can get complicated. Hopefully, this will make things easier. Let's see what the team does.

🌐 Website Improvements

Midjourney is also focusing on enhancing the user experience on their website. Based on user feedback, polishing touches are being added, including improved feedback on button clicks, tooltips, and the help page to prevent users from getting stuck. The team is gradually allowing random users to access the website without Discord, monitoring stats as more users are let in. The website will fully open once a certain level of polish is achieved and AB test stats look good.

Additionally, a chat room for help is being added to the website, and the team is experimenting with community features like trends and popular sref codes.

As a frontend developer I am really excited and stunned by how fast the team iterates and what UX solutions they come up with. Great job!

TL;DR;

🎨 Personalization Feature

• Released personalization feature, well-received

• Users need 200 ratings to personalize (midjourney.com/rank)

• Most users prefer personalization on

• Updating and improving personalization, debating update this week

• Plans for semi-personalization (e.g., demographic profiles, colorblind version)

• Investigating aesthetics and demographics

• Personalization is a work in progress, learning from users who dislike it

🔄 Blending Multiple Codes

• Updating code system to blend multiple codes

• Encouraging users to personalize, share, and blend codes

• Launch expected in a few hours after final QA

🚀 Upcoming Model Releases

• Training Midjourney 6.5, focusing on image quality and edge cases

• Midjourney Version 7 (V7) in progress, expected dramatic improvement over V6

🌐 Website Improvements

• Polishing website based on user testing

• Gradually allowing access without Discord, monitoring stats

• Adding chat room for help on the website

• Experimenting with community features like trends and popular codes

Stay tuned for more updates as Midjourney continues to innovate and enhance its platform!

If you’re not subscribed yet to my newsletter Imagine Weekly, I’d be thrilled to welcome you on board!