Just announced today, Midjourney V5 is here. Now, was it worth the wait? I'll let you decide. First and foremost, you can turn it on by going to your settings and picking V5, as you normally would if you were going to use V4 or Niji Journey.

💡
Mid-Journey, an AI-driven platform for generating images and artwork, recently held a three-and-a-half-hour live office hours event on their Discord channel. The event featured two major announcements and attracted nearly 2,000 attendees who were eager to learn more about Mid-Journey Version 5.

Introducing Mid-Journey Magazine

The first announcement during the office hours was the launch of Mid-Journey Magazine. This publication will showcase the diverse creativity of the Mid-Journey community. To receive the first issue for free, users can sign up at mag.midjourney.com and use the coupon code "subscriber." The magazine will feature a selection of artwork curated from the 10,000 most highly rated images, as well as interviews with Mid-Journey community members.

Mid-Journey Version 5: Release and Access

The second major announcement was the release of Mid-Journey Version 5. This new version is currently only available for paying subscribers, as the company does not have the computational power to offer it for free. To access Mid-Journey Version 5, paying members can either add "-V 5" to the end of any prompt or set it as their default version in their settings.

What's New in Midjourney V5?

Midjourney V5 brings several improvements, including a wider stylistic range, more responsiveness to prompting, higher image quality by two times, an improved dynamic range, and better performance with image prompting.

💡
Alpha Test and Future Modifications - It's essential to note that this is an alpha test, and the model will change. It will be significantly modified when the full release is out.

THE MOST IMPRESSIVE FEATURE - Improved Resolution and Photo Realism

Mid-Journey Version 5 boasts improved resolution, with all four images in a grid now upscaled by default. This means that users can view upscaled images almost instantaneously. However, right now it takes a bit longer to generate the initial four preview images. Additionally, the new version has significantly improved the realism of generated photographs, making images that are designed to look like photos appear more true-to-life.

Better Hands, Skin Textures, and Teeth

Version 5 has made strides in improving the generation of hands, skin textures, and teeth in generated images. Users will notice that hands are more accurate and skin textures now include realistic imperfections like blemishes, freckles, and moles. The teeth in smiling faces have also been greatly improved, avoiding the excessive toothiness that was common in earlier versions.

Improved Skin Textures: Mid-Journey version 5 has also focused on enhancing skin textures, making them more realistic. Generated images now feature skin blemishes, freckles, and even moles, adding to the lifelike appearance of the characters.

Teeth Generation Improvements: Another area where mid-Journey version 5 excels is teeth generation. In earlier versions, AI-generated images of people smiling often had more teeth than necessary, resulting in unnatural looks. This issue has been addressed in the latest version, making generated images with smiles more believable.

Better Generation of Hands: Hands have been a notorious challenge in AI-generated images. However, mid-Journey version 5 has made considerable improvements in the generation of hands, making them look more natural and accurate. While there might still be occasional oddities, the overall quality of hands in generated images has significantly improved.

💡
Comparing V4 and V5, at first glance, it still has that Midjourney look, but it does look different. You can really see that there's a difference in the dynamic range. Even with simple prompts, the texture of the makeup on the skin, especially around the eyes, on the cheekbones, and even on the mouth, has less of that plastic look.

Better Full Body Shots and Perspectives

One of the things Midjourney struggles with in stable diffusion platforms is when the subject's a little far in terms of perspective, especially with full body shots. However, in these examples, it's doing a much better job. The faces are more clearly defined, and the hands are not terrible either.

Revamped Prompting and Natural Language Processing

Mid-Journey Version 5 features improved natural language processing, which changes the way users should structure their prompts. Instead of using lists or single-word prompts, users should write their prompts like sentences to achieve better results. This new approach offers more powerful and detailed image generation, but users who prefer the older prompting style can still use Mid-Journey Version 4.

More Responsive to Inputs

In the Style Imprompting section for V5, they mention that it's more responsive to your inputs. This means that it's going to be more of a raw model that may require longer prompts to get the best out of it. Unlike with V4, you could put a short prompt in there, and it can look great. That's still the case, at least from my testing so far. But the benefit of being able to have a raw model to work with is that you'll be able to mold and shape it a little bit more exact to what you want to create.

New Features: Remix, Image Weights, and Aspect Ratios

Version 5 has introduced several new features, including better remix functionality for blending images, image weights for controlling the influence of images in text prompts, and broader aspect ratio options for generating images in various shapes and sizes.

Tileable Images and Backgrounds

Another exciting addition in Mid-Journey Version 5 is the ability to create tileable images by adding "-tile" to the end of a prompt. This feature allows users to generate images that can be seamlessly tiled, making it perfect for game designers and those looking to create unique backgrounds.

Improved Resolution and Upscaling

Mid-Journey version 5 has improved resolution for AI-generated images. Now, all four images in the grid are upscaled by default, which means that you can quickly view the high-resolution version of your desired image. This feature results in a faster and smoother experience when working with the platform.

Enhanced Photorealism

One of the significant changes in mid-Journey version 5 is the enhanced photorealism. Images designed to look like photographs will appear more like actual photos, making them more suitable for various applications, such as marketing materials, digital art, or concept development.

💡
With prompts like "fantasy art digital illustration," the details in the eyes and the metallic look in macro photography type of shots are very impressive. The intricate details, especially on mechanical ladybugs, show that the photorealism has improved quite a bit.

More Powerful and Detailed Prompts

One of the most exciting aspects of the V5 Algorithm launch was the improved prompting capabilities. Users could generate more accurate and compelling images based on their prompts, resulting in a more satisfying experience overall.

With the new version, the AI responds better to detailed and structured prompts. Instead of using a series of keywords separated by commas, users are encouraged to write their prompts like sentences, as the AI has improved natural language processing capabilities. This change allows for more control over the generated images and their details.

Enhanced Remix Feature

The remix feature, which allows blending two images together, has been improved in mid-Journey version 5. The new version provides more accurate and logical composite images, offering better results when blending images for various purposes.

Broader Aspect Ratio Options

Mid-Journey version 5 now supports a broader range of aspect ratios. Users can generate images in virtually any aspect ratio they can think of, allowing for greater flexibility and creativity in their projects.

Tileable Images

Another exciting addition to mid-Journey version 5 is the ability to create tileable images. By adding the "--tile" option to the prompt, the AI generates images that can be seamlessly arranged side by side, creating a continuous pattern. This feature can be particularly useful for game designers or those looking to create unique backgrounds.

Upcoming: Insights from Mid-Journey's Livestream

During the livestream, the Mid-Journey team shared valuable insights about their future plans, including:

  • The possibility of releasing an API for Mid-Journey.
  • The use of their own proprietary model instead of stable diffusion.
  • The expected release of Mid-Journey Version 6 within the next two months.

These developments suggest that we can expect rapid improvements and new features in the near future.

Limited Features: A Gripes with Midjourney

Importing, In-painting and Outpainting

I'm the kind of person where I want to have options like out painting and in painting  just like in other stable diffusion platforms such as PlaygroundAI, InvokeAI and LeonardoAI. The ability to bring in existing works and extend or replace certain aspects of the images with the power that Midjourney has would be amazing.

💡
Maybe I'm spoiled with all the progress that has been made with Stable Diffusion models and apps based on stable diffusion. Midjourney has always been great with the outcome and the quality that it delivers, but it's kind of limiting in terms of its features.

Text Generation Still Needs Improvement

Despite the impressive advancements in Mid-Journey Version 5, text generation remains a weak point. The AI-generated text is often nonsensical or unreadable. This means for things like mockups, labels, posters and logo's we're still going to have to reach for tools like Photoshop, Affinity or even Canva. Which i feel is such a shame given the power of the platform.

Deformed Text and Double Exposures

Some users have noticed large, deformed text appearing in their generated images. In addition, there have been reports of weird double exposures, causing image artifacts and inconsistencies. These issues are being actively investigated by the development team, who are committed to resolving them and providing a seamless experience for users.

Midjourney Censorship: NOT A Scalable Solution

This is a huge gripe for myself and many users. Midjourney has implemented a censorship system to filter out explicit content. However, the effectiveness of this scalable banning solution in handling various types of content, from legal adult material to gore images is debatable.

Midjourney's censorship system highlights the challenges of content filtering in an AI-powered art generator. As technology continues to advance, it remains to be seen how the platform will adapt its censorship system to better meet the needs of its diverse and creative userbase.

💡
I have a an opinion piece on Midjourney Censorship coming soon. Stay tuned...

Comparing Mid-Journey --V 4 and --V 5

One way to assess the improvements in Mid-Journey Version 5 is to compare it to its predecessor. By using the same prompts and comparing the generated images, we can see that Version 5 has made significant strides in producing more detailed and photorealistic visuals.

However, it's worth noting that Mid-Journey Version 4 may still produce better results when using less detailed prompts. This is because Version 5 is designed to work better with more specific and descriptive prompts.

Look-out for a more detailed comparison

Takeaway

The Future For Midjourney

As the MJ team continues to refine and improve v5, users can expect additional updates and enhancements in the future. By incorporating user feedback and addressing reported bugs, the team is dedicated to ensuring that the new "live" v5 offers an enjoyable and fulfilling experience for all users.

With an API on the horizon and Mid-Journey Version 6 expected within two months, the future looks bright for this powerful AI tool. As it continues to evolve, we can expect even more impressive image generation capabilities and new features to explore.

Was Midjourney V5 Worth the Wait?

Mid-Journey Version 5 offers significant improvements in image generation, particularly for detailed and photorealistic visuals. But it still does not offer the flexibility that many other tools in the generative art space offers.

💡
I was considering not resubscribing for Midjourney since I hardly use it in comparison with other Stable Diffusion models. But I'll let it run to see how --V5 works out..

There are still some limitations with text generation and censorships and must have features to compete with the other art generators on the market at the moment. But the tool's rapid development promises exciting new features and capabilities in the near future.

So, what do you guys think? Let me know in the comments below. Is it worth the wait to you, or do you prefer other stable diffusion platforms?

Share this post