🚀 Midjourney Office Hours Highlights: June 25th

Midjourney is optimizing its video infrastructure to cut high cloud costs and improve performance, especially in Relax Mode, which faces usage-based restrictions due to financial strain. Operating without investors, the company relies on community subscriptions—80% of revenue funds model training and operations. New video tools like a Video Upscaler, Turbo Mode, and Start/End Frame Generation are in development, alongside major image model updates (v7.1 and early v8). Platform growth remains steady but capped by its paid model, with new user incentives like fast-hour rewards. Looking ahead, Midjourney plans deeper image-video integration, larger model training via partnerships, mobile feature rollouts, and the reveal of two secret projects.

📹 Video Infrastructure & Costs

Midjourney's recent office hours shed light on the company's ongoing efforts to optimize video generation infrastructure. The team is focusing on reducing cloud infrastructure costs, as the video model is currently expensive to operate. Efforts throughout the month are dedicated to server optimizations, which, if successful, could result in faster generation times, reduced costs, and enhanced performance in Relax mode.

🕒 Relax Mode Changes

Relax mode, originally designed to utilize excess capacity, is facing changes due to its heavy use by the top 1% of users, which is causing financial losses. The Midjourney team is implementing changes such as increased wait times, more variable wait times, or adjusted concurrent job limits to address this issue. The previously available capacity has diminished, necessitating these adjustments.

💼 Business Model & Funding

Midjourney operates on a community-funded model, with approximately 80% of subscription revenue going towards running and training models. Unlike many tech companies, Midjourney does not rely on investors. Despite training their current video model with significantly less compute than giants like Google, they are exploring partnerships with other AI labs to access additional compute resources. Higher-tier subscriptions are crucial for funding the training of larger models.

🎥 Video Features in Development

Exciting video features are in the pipeline. The team is working on a Video Upscaler, aiming to reduce artifacts, and developing Start/End Frame Generation to enable smooth animations and looping videos. A Turbo Mode for Video is also being developed, offering much faster generation at a higher cost for professional users. Additionally, an update using reinforcement learning promises quality improvements based on user data.

🖼️ Image Model Updates

The image model is also seeing significant updates. Version 7.1 aims to integrate video and image models, although the release bar is set high. Meanwhile, Version 8 is already in early training stages, with ambitious data preparation underway. The Niji Anime Model is on hold due to unsatisfactory quality, but potential mood board performance improvements are under testing.

🌐 Platform Growth & Stability

The recent major platform launch was the smoothest in the company's history, despite massive changes. The team has grown to around 60 people, enabling larger releases. While new user growth remains steady, it is limited by the paid-only model. A new video ranking feature now allows users to earn extra fast hours, further enhancing the platform's appeal.

🔮 Future Outlook

Looking ahead, Midjourney plans to increasingly combine image and video capabilities within its models. Partnerships may enable the training of much larger models, and mobile features for video and images are being explored. Two secret projects are nearing announcement, promising exciting developments. With a larger team, more frequent major releases are expected, driving the company's growth and innovation.

TL;DR

📹 Video Infrastructure & Costs: Efforts are underway to reduce video generation costs through server optimizations, aiming for faster and cheaper production.

🕒 Relax Mode Changes: Changes in Relax mode are being implemented due to heavy usage by top users, impacting financials.

💼 Business Model & Funding: 80% of revenue is reinvested in models; company is community-funded and exploring AI lab partnerships for resources.

🎥 Video Features in Development: New features like Video Upscaler, Start/End Frame Generation, and Turbo Mode are in development to enhance video capabilities.

🖼️ Image Model Updates: Version 7.1 and 8 are in progress, with ambitious integration goals and testing for mood board improvements.

🌐 Platform Growth & Stability: Smooth platform launch with a growing team; new user growth steady but limited by paid model.

🔮 Future Outlook: Combining image and video capabilities, exploring mobile features, and two secret projects nearing announcement.

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!