What's New with Midjourney? Key Updates from the Latest Office Hours

Curious about what the Midjourney team is working on? The recent office hours provided some insights into current challenges, upcoming features, and future plans. From infrastructure improvements to new tools and development philosophies, here’s a look at what was discussed.

Behind the Scenes: Infrastructure and Website

Midjourney has faced some database issues and outages recently. The team is investing more in their infrastructure to prevent these problems. They said it's tough to avoid issues completely with a system as large as Midjourney, though.

Some user pictures from early April are missing, but this was a manual issue and the team plans to restore them soon. They are also working to make the database handle large numbers of images better, so users with thousands of pictures can load and search instantly on the new website.

Finding great community images will also get easier. The team is developing better ways to explore user-created artwork and has promising early results. This feature might appear in the next version of the Midjourney website.

Upcoming Features: In-painting and Aesthetics

The in-painting feature might be released soon. The head of Midjourney, David Holz, mentioned it could come out this week, though he isn't fully happy with it yet. He noted that you often only see the best results for in-painting on other platforms, not the errors.

There's also consideration for an aesthetic update. If the in-painting feature releases without this update, it will likely stay as version 5.2. If the aesthetic improvements come with it, they might call it version 5.3.

The team is exploring if this next version could also be more accurate to user text prompts. This is not a guarantee, just something they are looking into.

The Development Challenge

Deciding what to focus on is a challenge. Making the core experience better, like improving image prompting for everyone, is seen as crucial. This might take priority over adding more advanced features like panning right now.

Trying to do a bit of everything rather than focusing on one main goal is a strategy decision. David hinted this isn't his preferred approach, but it's what they are doing for now.

Developing new features takes time. Research might only work 10-30% of the time, and experiments can take a week each. It's hard to know what will actually succeed. Shipping a complete version also requires significant team effort, so they prefer releasing larger updates over small, frequent ones.

Version 6 and Beyond

Version 6 is currently behind standards or hopes. The team might hold off releasing it until all features are ready. Or they might set a release date and put out whatever they have at that time. They feel improving prompt understanding is most important to most users, and this is a goal for version 6.

Looking much further ahead, there's potential for another big visual leap like the one seen from version 3 to 4. This is unlikely to be version 6 but could happen in version 7 or later.

There was a brief mention that 3D rendering is getting better, and video generation is improving. However, there is no way to guess exact timelines for these features.

Understanding User Preferences

Midjourney users have different preferences for art. The biggest differences seem to be based on age and gender. Location also plays a role; for example, users in Japan have different tastes than those in other countries like Germany or the UK.

This data leads to questions about how Midjourney should respond. Providing more variety in larger image batches is one idea. Another idea discussed previously was a personal style code based on your art preferences. However, the speaker hoped Midjourney avoids tuning the main algorithm based on grouping users by demographics. They prefer everyone playing by the same general rules.

Managing vast libraries and experimenting with different styles based on these insights can be time-consuming. A tool that manages your creations and allows you to easily apply different styles or explore variations could be very helpful. Consider how the Midjourney Automation Suite from TitanXT might streamline your workflow.

Exploring Styles and Prompts

The speaker shared a prompt used for generating art in their video: Subject imagined by Yves Tanguy, Ellsworth Kelly, and Nicholas Poussin. They also used `--stylize 400` and `--chaos 20`.

Experimenting with artist names, styles, and parameters is key to getting unique results. Keeping track of successful combinations and managing different prompt variations can become complex quickly.

If you find yourself creating many images and needing a better way to organize, manage, and replicate successful styles or parameters, check out the Midjourney Automation Suite from TitanXT. It's designed to help you handle these tasks efficiently.

In Summary

Midjourney is actively working on improving its core infrastructure and website, preparing for features like in-painting and potential aesthetic updates, and grappling with the challenges of development and resource allocation. They continue to aim for better prompt understanding in future versions and are aware of the diverse preferences among users. While specific timelines remain uncertain, the team is focused on building a more robust and capable platform.

Staying on top of these updates and managing your Midjourney workflow can be simplified. Explore tools like the Midjourney Automation Suite to enhance your creative process.