Midjourney V6 is out; it’s the most powerful AI image generator so far

There’s been a buzz in the community that Midjourney was about to release V6. And as of last night, it’s out! Midjourney has opened the alpha version for community testing over the winter break, starting from yesterday. And judging from the results so far, it’s the most powerful AI image generator to date.

While Midjourney is supposed to move to a dedicated website soon, the V6 is available through Discord just like the previous ones. You can simply select ‘V6’ from the dropdown menu under /settings or type –v 6 after their prompt in Midjourney’s interface.

What’s new in V6?

Midjourney announced the news of the V6 release on Discord, and here’s what you can expect:

• Enhanced Prompt Accuracy and Length: The V6 model excels in following prompts more accurately and supports longer prompt inputs.
• Improved Coherence and Knowledge: This version offers a significant upgrade in coherence and the model’s inherent knowledge base.
• Advanced Image Prompting and Remix Capabilities: The V6 model has improved in terms of image prompting and the ability to remix images.
• Text Drawing Capability: Users can now incorporate minor text elements in their images. This is achieved by writing the desired text in quotations and adjusting the style settings.
• Upgraded Upscalers: The new upscalers in V6 come in ‘subtle’ and ‘creative’ modes, enhancing image resolution by 2x.

Probably the biggest news is Midjourney’s possibility to write. So far, I’ve only seen Ideogram write text accurately (after a few attempts though). But judging from Midjourney V6 test images, the two tools are now a serious competition in this regard. To be fair, it needs some more refinement, but let’s remember that it’s still in the alpha stage.

Supported and upcoming features

The launch version of V6 supports features we’ve already seen before: –ar, –chaos, –weird, –tile, –stylize, –style raw, Vary (subtle), Vary (strong), Remix, /blend, and /describe. The upcoming features include Pan, Zoom, Vary (region), /tune, and a new V6 version of /describe, which is expected to roll out within the next month.

Prompting: Forget everything you knew

Prompting in V6 differs significantly from V5, according to the developers. In fact, they claim that you’ll need to “relearn” the art of prompting. The new model is highly sensitive to prompts, and it needs clear and explicit instructions for optimal results. So, avoid vague descriptors and focus on being as specific as possible about your requirements.

Midjourney encourages community collaboration, inviting you to share your experiences and learnings in the prompt-faqs to master the V6’s capabilities. You are also invited to share your creations in the v6-showcase on Discord. Finally, you can give your feedback in the ideas-and-features section. The website still isn’t up at the moment of writing this, but we expect it to launch any day now, so stay tuned!