Midjourney V6 is here with in-image text and completely overhauled prompting

Call it a holiday present: Midjourney version 6, the latest and greatest iteration of the popular image generation AI model from the research collective of the same name founded by David Holz, dropped last night as an alpha release — and already, some power users are ecstatic over the improvements it brings. VentureBeat uses Midjourney and other AI art tools to generate article imagery.

Among those new features are drastically improved and more realistic, highly detailed images, and the ability to have the model generate legible text within images, something that had eluded Midjourney since its release in 2022 even as other rival AI image generators such as OpenAI’s DALL-E 3 and Ideogram had launched this type of feature.

“This model can generate much more realistic imagery than anything we’ve released before,” wrote Holz in a message posted in the Midjourney Discord server, which has over 17 million members. Holz said V6 was actually the “third model trained from scratch on our AI superclusters” and took nine months to develop.

How to enable MJ V6?

The update won’t take effect for users by default — at least, it didn’t for me. You’ll need to type in the slash command “/settings” in the Midjourney Discord server or in a direct message (DM) to the Midjourney bot and then use the dropdown menu at the top to select V6. Or, you can do it the old school way and manually type “–v 6” after your prompts.

What’s new in MJ V6?

Specifically, Holz called out several new features, including:

• “Much more accurate prompt following as well as longer prompts
• Improved coherence, and model knowledge
• Improved image prompting and remix
• Minor text drawing ability (you must write your text in “quotations” and --style raw or lower --stylize values may help)

/imagine a photo of the text "Hello World!" written with a marker on a sticky note --ar 16:9 --v 6

• Improved upscalers, with both 'subtle‘ and 'creative‘ modes (increases resolution by 2x)”

New prompting methods encouraged

The founder and leader of the Midjourney project also clarified that an entirely new prompting method had been developed.

Midjourney’s prompting — how users generate images by typing in specific text descriptions and key words into the Discord server or alpha version of the website — had long been somewhat esoteric and technical, with users sharing examples of techniques that had worked well for them on social media, such as including camera names (e.g. Leica M11), film stock (35mm), and resolution (8k), to get high quality, photorealistic or cinematic results out of the AI model.

Yet Holz was clear in his Discord post stating that these types of prompting tricks would no longer result in the type of results users desired. “You will need to re-learn how to prompt,” he wrote.

• “Prompting with V6 is significantly different than V5. You will need to ‘relearn’ how to prompt.
• V6 is MUCH more sensitive to your prompt. Avoid ‘junk’ like “award winning, photorealistic, 4k, 8k”
• Be explicit about what you want. It may be less vibey but if you are explicit it’s now MUCH better at understanding you.
• If you want something more photographic / less opinionated / more literal you should probably default to using --style raw
• Lower values of --stylize (default 100) may have better prompt understanding while higher values (up to 1000) may have better aesthetics
• Please chat with each other in ⁠prompt-chat to figure out how to use v6.“

Initial results

I tested MJ V6 myself briefly this morning before writing this article and I’m sorry to say that so far, for me at least, the update has been a little underwhelming. While I definitely observed increased detail and more photorealistic generations, the results weren’t so different enough that I would have been able to tell just by looking at a V5.2 or V6 generation side-by-side.

I was, however, impressed with the lighting effects and reflection details that are able to be generated.

Other avid users including horror director and digital artist Chris Perna have begun testing and posting incredibly vivid, richly detailed results generated by MJ V6 on Instagram and other social media sites. And the early examples of text generation look really promising.

And as Holz noted in his Discord message announcing V6, the new model “is an alpha test. Things will change frequently and without notice…It will significantly change as we take V6 to full release…V6 isn’t the final step, but we hope you all feel the progression of something profound that deeply intertwines with the powers of our collective imaginations.”

In addition, V6 is currently missing some features found on V5.2 including pan left and right and zoom out, but Holz said these would be coming in later updates to V6.

The updates show Midjourney continues to progress its model — considered by many to be the preeminent and highest quality, as well as most creative — AI art generator currently available, retaining its leadership even as it faces challenges from competitors using their own in-house models or the popular open-source Stable Diffusion model, which relies on a popular underlying AI technology called “diffusion,” where algorithms are trained to recreate images from visual “noise.”

Meanwhile, Midjourney and other diffusion-based AI art generators are facing class action litigation for copyright infringement by artists who accuse them of training on their publicly posted work without affirmative consent or compensation, though early indications suggest the AI art generators have a strong “fair use” defense.