🚀 Midjourney Office Hours Highlights: April 23rd

🚀 Features & Testing: Major Experiments in the Works

The latest Office Hours saw Midjourney’s team testing new default image settings, with internal guides and moderators reporting a clear preference over prior versions. A new experimental “less-conservative” tuning mode is also being trialed, significantly boosting visual quality, but at the expense of precise prompt adherence. The team is undecided on when, or if, this mode will see a public release.

🖼️ Omni-Reference & Personalization: Progress and Roadblocks

Omni-Reference—the system enabling image-based prompts—has proven solid for single-image input but stumbles with scenes involving multiple subjects, such as two faces or unconventional object arrangements. Meanwhile, prototypes for personalization around aesthetics and style are advancing, though not yet ready for public trials due to the effort needed for thorough evaluation.

🛠️ Release Strategy & Quality Balance: Speed Versus Polish

There’s an internal push to ship updates more quickly, even if they aren’t perfect, to keep user creativity at the forefront. This has sparked ongoing debate about finding the right balance between rapid releases and product polish, particularly after last week’s swift rollout met with mixed community reactions. A recent holiday weekend also caused a two-day slip in the release schedule.

🎨 Image Settings & Personalization Roadmap: Quality and the Future

The conservative image quality update is considered “good enough” and poised for immediate launch, although a short delay is being considered to test a more ambitious variant. In the longer run, Midjourney plans regular, community-driven personalization updates, with new rating tools to help steer improvements. Ambitiously, the V7 release could deliver up to a tenfold leap in output quality, but robust tools are needed to reliably measure such jumps.

🖼️ Omni-Reference Status: Almost Ready, With Fixes Pending

Omni-Reference is now delivering around 80–90% successful results, but about 10% of outputs still show unwanted “burnt” artifacts. Engineering work is underway to address these, and the team is developing a more nuanced moderation framework to balance creative freedom with safety. If all goes well, Omni-Reference could see public release within the next week.

⚡ Fast Mode & “q 4” Engine: Speed with Minor Tradeoffs

The fast-render “q 4” engine is showing only a slight (about 2%) dip in image quality for some edge cases. The team aims to launch it alongside the Omni-Reference feature, possibly as soon as next week.

📹 Video Model Progress: Early Wins and Next Steps

Midjourney’s video model has produced its first viable results, impressing with unexpectedly high quality for the cost. The plan is to make this feature affordable for existing subscribers, with aesthetic consistency that outpaces current market options, though a few quirks remain. More training and optimization are needed, so the earliest possible public preview would be next month.

🛤️ Upcoming Milestones: What’s Next on the Roadmap

The V7 update, integrating Omni-Reference, refreshed aesthetics, and Fast Mode, is on track for release in the coming weeks. After that, attention will likely turn to V8 rather than 3-D generation, as the latter doesn’t offer as much near-term potential. Two competing designs for an enhanced style-reference system are also under consideration, part of a broader push toward “aesthetic singularity”—the ambition to massively expand creative style options within a year.

🏗️ Tooling, UI, & Infrastructure: Under-the-Hood Improvements

Continued improvements are being made to the in-app editor to ensure a smoother experience for all images. New website workflows are being explored to collect user feedback faster, outpacing traditional guide/mod review cycles. Behind the scenes, “collective intelligence” projects and hardware upgrades are progressing, with the first public details expected in about six months.

🌏 Community & Commercial Considerations: Keeping Users at the Core

Midjourney remains committed to prioritizing fun and creative freedom for its users, even as some subscribers push for more commercial-grade results. There’s an active effort to broaden the testing pool, allowing everyday users to influence image quality alongside internal guides and moderators.

TL;DR

🚀 Testing: New default image settings and experimental tuning modes; Omni-Reference strong on singles, tricky on complex scenes.

🛠️ Release: Faster updates prioritized, but debate on polish; schedule delayed slightly by holiday.

🎨 Personalization: Style & aesthetic personalization in the works; V7 could be a huge quality leap.

🖼️ Image: Conservative quality update nearly ready; Omni-Reference aiming for next week after artifact fixes.

📹 Video: Early video model promising, public preview possibly next month.

🏗️ Infrastructure: Editor and feedback tools improving; hardware and “collective intelligence” upgrades coming.

🌏 Community: User creativity prioritized; plans to expand test pools for more inclusive quality input.

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

If you want to support me, feel free to buy me a coffee ☕️ 

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