7 mins read

Midjourney V6 Announced: Sharper, More Detailed, and Realistic

Better at understanding long and complex prompt instructions, Midjourney V6 can generate text accurately and produce extremely realistic images. It’s a bit slower but much more powerful — now in alpha test.


If you haven’t heard about it yet, Midjourney is one of the first AI image generators available, transforming text prompts into pictures, similar to DALL-E 3 and Stable Diffusion. It enables users to create landscapes, portraits, and abstract artwork with just a few words. Its unique style, a mix of realism and cinematic flair, has been generally much appreciated by the community. Anyway, Midjourney can also generate a wide array of art styles, including surreal, expressive, fantasy, abstract, photorealistic, anime, pop art, graffiti, pixel art, and traditional painting.

Midjourney offers different subscription tiers, and images are generated using Discord. This might be a downside for some, as a more user-friendly interface would likely be preferable.

Midjourney V6: What’s New?

Midjourney has just rolled out its latest version, V6, and it’s available for testing on Discord as an alpha version, so things might change quickly and without notice. To try it out, all you need to do is select ‘V6’ in the /settings dropdown menu or simply type --v 6 after your prompt.

Selecting midjourney V6 on Discord

So, what’s new in this release? Let’s see what the announcement says. First off, your prompts can now be longer and more intricate, and Midjourney V6 will interpret them with remarkable accuracy. Additionally, the model’s understanding has significantly improved, meaning your creations will be more coherent and lifelike. Currently V6 is slower and more expensive compared to the previous version V5, but will get faster with future optimizations.

Accurate text

Midjourney V6 brings in improved image prompting and remix capabilities. Now, it even lets you add text to your images, something that currently only DALLE-3 seems to be doing very well, whereas Stable Diffusion still struggles (but it’s not impossible either).

Supported features

Moreover, there are a bunch of cool features supported at launch, including --ar, --chaos, --weird, --tile, and more.

  •     –ar
  •     –chaos
  •     –weird
  •     –tile
  •     –stylize
  •     –style raw 
  •     Vary (subtle)
  •     Vary (strong) 
  •     Remix
  •     /blend

Additional features like Pan, Zoom, Vary (region), /tune and a new version of /describe are still in the works and will be added soon.

Prompt understanding

As stated in the announcement, apparently Midjourney V6 requires a bit of relearning when it comes to prompting. It’s super sensitive to your instructions, so being clear and explicit is key. Want something more realistic? Use –-style raw. And for the artists out there, playing with --stylize levels can alter the aesthetics of your images.

midjourney v6 prompt understanding

editorial shot, shot with agfa vista, a blonde woman and a brunette man sitting in a high-end cocktail bar, 16th floor, night, golden / grey colors, cinematic lighting, Frankfurt am Main, minimalistic style, 8k, --style raw --ar 16:9 --v 6.0 – Variations (Strong) by @strike (fast)

Lower values of –stylize (default 100) may have better prompt understanding while higher values (up to 1000) may have better aesthetics. It also finally possible avoid terms like award winning, photorealistic, 4k, 8k, which have been overused since the first model was introduced.

Currently in Alpha test

But, as with all alpha tests, remember that things might change without notice. Midjourney V6 is a bit slower and pricier than V5, but it’s set to become faster with further optimization.

Midjourney also emphasizes community standards, especially since V6 can create ultra-realistic imagery. The team has enhanced moderation, reminding everyone to create responsibly and respectfully.

Lastly, a bit of background on V6: it’s Midjourney’s third model, developed from scratch over nine months using their AI superclusters.

So, there you have it! Midjourney V6 is here, offering an alternative to DALLE and Stable Diffusion. Test it out and share your creations in the #v6-showcase channel.

Key Points

  1. Enhanced Prompt Accuracy: Midjourney V6 promises more accurate following of longer and more detailed prompts, improving the relevance of the generated images to the user’s request.
  2. Improved Coherence and Knowledge: The model now demonstrates better coherence in the images it produces, alongside an upgrade in its ‘knowledge’ or understanding of the world, likely leading to more realistic and contextually accurate images.
  3. Advanced Image Prompting and Remix: These features allow for more creative control over the image generation process, including the ability to remix and modify existing images.
  4. Text Drawing Capability: Users can now include text in their images with more precision, offering new avenues for creative expression.
  5. Improved Upscalers: The new upscaling options, ‘subtle’ and ‘creative’, enhance the resolution and quality of images, providing clearer and more detailed results.
  6. New and Pending Features: The announcement details various supported and upcoming features, indicating a significant expansion of the tool’s capabilities.
  7. Prompting and Style Changes: V6 requires users to adapt their prompting techniques. It emphasizes clarity and specificity in prompts, offering different aesthetic outcomes based on the stylization levels.
  8. Community and Testing Focus: The release of V6 is positioned as an alpha test, inviting the Midjourney community to actively engage, experiment, and provide feedback.

Comparing Midjourney, DALL-E 3, and Stable Diffusion

Midjourney

Known for its artistic flair and ability to produce highly stylized and unique images. Midjourney’s offers many functionalities such us upscaling, image variations and specific arguments that make easier to control your output. Currently it’s only available on discord.

DALL-E 3

DALL-E 3 is celebrated for its ability to understand and generate images from complex prompts. It’s particularly adept at creating images that blend concepts in surreal or imaginative ways while maintaining a high level of detail. Currently, It’s not really good at generating realistic pictures.

Stable Diffusion

This model stands out for being open source (with some conditions) and the ability to run on consumer-grade hardware. It’s particularly useful for users who want to generate high-quality images, directly from their PC, without requiring substantial computational resources. If you want to get started with Stable Diffusion, here you can find a set of tutorials.


First pictures from V6

These images create by users from the #v6-showcase channels, and I have to say some of them are really impressive.

If you also want to get started with Midjourney V6, discover their website and get a subscription using a Discord account.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.