Call it a holiday present – Midjourney Version 6, the newest iteration of the renowned image generation AI model, has just been released. Developed by David Holz and his research collective, this alpha release has already garnered praise from power users for its significant improvements. Midjourney, utilized by VentureBeat and other AI art tools, now offers highly detailed and realistic images, as well as the ability to generate legible text within images. This feature has long eluded Midjourney, making it a notable addition to the AI image generation landscape.
Enhanced Realism and Detail
“This model can generate much more realistic imagery than anything we’ve released before,” wrote David Holz in a message posted on the Midjourney Discord server. With over 17 million members, the server serves as a hub for enthusiasts. Holz revealed that Version 6 is actually the third model trained entirely from scratch on their AI superclusters. The development process took an impressive nine months, culminating in the release of this highly anticipated update.
To access Version 6, users will need to take an extra step. By typing in the command “/settings” on the Midjourney Discord server or through a direct message to the Midjourney bot, users can select Version 6 from the dropdown menu at the top. Alternatively, users may manually type “–v 6” after their prompts, the old-fashioned way.
Holz highlighted several notable new features as part of Version 6, including the ability to generate a photo of the text “Hello World!” written with a marker on a sticky note. Users have the flexibility to choose aspect ratios and further personalize their creations.
A New Prompting Method
In addition to these updates, Midjourney has also introduced an entirely new prompting method. Previously, users had employed esoteric and technical techniques to generate the desired images. Examples shared on social media often involved including specific camera names, film stock details, and resolution specifications. However, Holz emphasized that these tricks no longer yield the desired results. Users will need to adapt and “re-learn” how to prompt the AI model effectively.
While I personally found the initial results of Version 6 underwhelming, with the improvements not immediately apparent when compared directly to previous versions, other users like horror director and digital artist Chris Perna have already started testing and posting stunning results on social media platforms. The text generation capabilities of Version 6, in particular, show immense promise.
Holz acknowledged that Version 6 is still in its alpha stage and subject to frequent changes. He assured users that the model will continue to evolve, drawing on the collective imaginations of the Midjourney community. Future updates to Version 6 will include missing features from previous versions such as pan left and right and zoom out.
As Midjourney progresses in refining its AI model, it remains the leading AI art generator in terms of quality and creativity. While competitors may have their own in-house models or rely on the open-source Stable Diffusion model, Midjourney continues to maintain its position at the forefront. However, like other diffusion-based AI art generators, Midjourney is currently facing class action litigation for alleged copyright infringement. Artists claim that their publicly posted work was used without consent or compensation. It is worth noting, though, that Midjourney and other AI art generators may have a strong “fair use” defense.
