During recent "Office Hours," Midjourney founder and CEO David Holz personally shared updates on current projects and answered community questions via Discord.
After rapid progress in a short time, the frequency of Midjourney's new model releases has noticeably slowed. The startup's most recent release was version 6.1 in July, following a long wait. Version 7, originally planned for release before summer, now has a new timeline: it should be ready in one to two months.
Instead of rushing new models, the developers are focusing on making the technology more accessible and the tool more useful for professional daily use. For many, the requirement to use Discord was initially off-putting. Now, a web interface is available to all users, including those wanting to try the image AI for free.
While the turbo mode introduced in April has already significantly sped up image generation, users will soon be able to create eight images at once. Since the beta release, this has been limited to four images per run, similar to other image generators like Ideogram.
Midjourney is also currently working on an image editor that allows users to edit external images, such as photos, using inpainting and outpainting functions. It's unclear whether this will be integrated into the existing platform or become a separate product. No timeline was given for this feature. Until then, Photoshop with Generative Fill will likely remain the best option for this use case.
Midjourney wants to immerse you in images
In a more ambitious project, Midjourney is developing a 3D system that allows users to virtually immerse themselves in AI images. Instead of using polygon rendering, NeRFs, or Gaussian splatting, they're creating a new "NeRF-like" format. The team is led by someone who played a key role in the invention of NeRFs, leading some to speculate that it could be former Nvidia researcher Alex Evans, who joined Midjourney last year.
Personalization is also at the top of Midjourney's list of new features. Based on user feedback, the tool aims to learn more about user preferences and deliver more individualized results. Midjourney recently enabled this feature for the Niji model, which specializes in anime characters.
Midjourney seems to have a full plate of developments. Beyond its image generation results, the company also plans to tackle video generation. In addition to software, Midjourney is investing in hardware development as well.
