🚀 Midjourney Office Hours Highlights: February 19th

Midjourney, is gearing up to introduce a series of significant advancements that could transform digital content creation. With the development of their Version 7 (V7) model and explorations into video and real-time 3D, Midjourney is pushing the boundaries of what's possible in the digital art space. This post provides a straightforward look at the upcoming innovations and what they mean for the future of creative technology.

For a quick overview, head straight to the TL;DR; section otherwise keep on reading.

Extended Recap

🎨 V7 Image Model

MidJourney is working on two versions of the V7 image model: a fast model and a slow model. The fast model aims to generate images in 0.5 to 2 seconds, though it may sacrifice some quality. In contrast, the slow model promises higher quality and takes 20-50% less time than its predecessor, V6. Despite facing initial challenges with resolution settings and learning rates, the team is focused on personalization, prompt accuracy, and aesthetics. Omni References, aimed at improving character consistency, are in progress and could be revolutionary. An image rating exercise may follow the release to refine the model further.

🌐 Website Features & Improvements

Several enhancements are underway to improve user experience. These include a Batch Mode for processing multiple images, Real-Time Mode for live image refinement, and Workspace & Folders for better organization. Additionally, Region Selection and Layer Tools will offer users more control in the editor.

🎥 Video Model

MidJourney is considering developing a fully in-house video model to keep costs low, aiming for a $10/month price point. The initial focus will be on image-to-video transformation and editing tools, with community input potentially guiding improvements. The team is mindful of training costs and ensuring a smooth user experience at scale.

🛠️ Challenges in V7 Training & Model Design

The training of the V7 model faced early setbacks due to misconfigured resolution and learning rates. While adjustments have improved performance, the slow model requires additional training. The team is balancing speed and quality, with users potentially choosing between the fast and slow models based on their needs.

🌌 Future Directions: AI in Video & 3D Models

MidJourney is exploring multiple approaches for video generation, including in-house models and partnerships. The goal is to create an affordable model not restricted by high pricing tiers. Initial versions will focus on simple generation and scene extensions, with future iterations allowing for camera movement and advanced editing. In 3D modeling, the team is investigating real-time AI-generated 3D spaces, contingent on model speed improvements.

🤝 Community Involvement & Feature Requests

Community feedback will play a crucial role in fine-tuning the V7 and video models. Users have requested improvements in tiling, UI mockups, and tool selection for better control. Workspaces and folders will be introduced for enhanced organization, though vector graphics models are not currently planned.

🌍 MidJourney’s Market Position & Industry Commentary

MidJourney believes that while Google has significant AI talent, execution remains a challenge. They see startups as key to innovation, though big tech consolidation has hindered progress. China is excelling in video AI but is less focused on image generation. MidJourney aims to provide affordable, high-quality AI tools to fuel imagination and creativity.

💭 Personal & Philosophical Takes

MidJourney emphasizes that AI should enhance creative ambitions rather than replace artistic effort. They encourage using AI for faster ideation, allowing people to tackle more complex projects. The team is skeptical about quantum computing's general-purpose applications but optimistic about Gen Alpha's potential to reject social media toxicity.

🔮 Miscellaneous & Future Plans

MidJourney is exploring future physical products, one of which could save lives, and another that might be an iPod-level innovation. They draw inspiration from Star Trek's explorer narrative and recommend the music track "Final Form" by Henry Green. The company is hiring web designers, AI researchers, cloud engineers, and data experts.

📅 Estimated Release Timeline

Fast V7: 1-4 weeks (finalizing aesthetic improvements) Slow V7: 1-3 weeks more training needed Video Model: Likely after both V7 models (mid-year) 3D Tools: Experimental, dependent on V7 speed improvements

TL;DR

🎨 V7 Image Model: Fast & slow versions; focus on quality and speed. Omni References for character consistency.

🌐 Website Features: Batch Mode, Real-Time Mode, Workspace & Folders.

🎥 Video Model: In-house model for affordability; image-to-video tools.

🛠️ Challenges & Design: Training issues resolved; balancing speed and quality.

🌌 Future Directions: Affordable video & 3D models; real-time 3D spaces.

🤝 Community & Features: Image rating, UI improvements, workspaces.

🌍 Market Position Startups drive innovation; focus on affordable AI tools.

🔮 Future Plans: Physical products; hiring focus; music recommendation.

📅 Release Timeline: Fast V7: 1-4 weeks; Slow V7: 1-3 weeks more training.

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!