Join our daily and weekly newsletters to stay updated with the latest AI industry news and exclusive content.
Think of it as an early holiday gift: Midjourney version 6, the newest iteration of the popular image generation AI model, was released last night as an alpha version. This new release has already thrilled many advanced users with the improvements it offers. VentureBeat often uses Midjourney and other AI tools for creating article images.
The latest version includes features such as more realistic and detailed images and the ability to generate readable text within images, a feature that had previously been missing. Unlike other AI image generators like OpenAI’s DALL-E 3 and Ideogram, Midjourney had lagged in this aspect.
David Holz, the founder of Midjourney, announced these updates on their Discord server, which has over 17 million members. He mentioned that version 6 is actually the third model trained from scratch and took nine months to develop.
To enable Midjourney V6, users need to type the command “/settings” in the Midjourney Discord server or send a direct message to the Midjourney bot. Then, use the dropdown menu to select V6. Alternatively, you can manually type “–v 6” after your prompts.
Some notable features of V6 include:
– More accurate prompt following and support for longer prompts
– Improved coherence and model knowledge
– Enhanced image prompting and remix capabilities
– The ability to draw minor text within images
– Better upscalers with ‘subtle’ and ‘creative’ modes that increase resolution by 2x
Holz also highlighted a new method for creating prompts, suggesting that previous techniques, such as including camera names or specific film stock, would no longer yield the desired results. He emphasized the need to be explicit and avoid generic terms like “award-winning” or “4k.” Using “–style raw” can help achieve more photographic results, and lower values of “–stylize” might lead to better prompt understanding.
Although I briefly tested V6 and found the improvements subtle, especially when comparing the details and photorealism with V5.2, the lighting effects and reflection details were impressive. Other users, including horror director and digital artist Chris Perna, have shared excellent results on social media.
Some features present in V5.2, like pan left and right and zoom out, are still missing but are expected to be included in future updates.
Midjourney continues to enhance its model, which remains a leading and highly creative AI art generator despite facing competition from other in-house models and open-source alternatives like Stable Diffusion. However, there are ongoing legal challenges alleging copyright infringement due to training on publicly posted works without consent or compensation, though early signs suggest a strong “fair use” defense for the AI generators.