Midjourney V5 is here!
A recent breakthrough in artificial intelligence has been making waves: the highly anticipated Midjourney V5 has finally been released. The independent research lab responsible for this development has unveiled the latest iteration of their renowned AI art generator. Many have already dubbed it “a world of photorealistic wonder” for its ability to create awe-inspiring images based on text prompts. As we delve deeper into the new capabilities of this updated model, we will discover how it has enhanced language understanding, accuracy, and stylistic flexibility.
The Midjourney V5 is the second deep-learning model developed by the team and has been in development for the past five months. It boasts a completely different neural architecture and novel aesthetic techniques than its predecessor. The developers describe it as “newly trained, bigger-brained, that it knows more, understands more, and listens better.” Naturally, we wanted to experience these improvements firsthand. We were pleasantly surprised by the impressive results it produced.
How to Use Midjourney V5
Assuming you have familiarized yourself with Midjourney, you may wonder how to integrate the new V5 model into your creative workflow. Fortunately, there are two methods to accomplish this. First, you can enter “/settings” in any #newbies Discord channel and select “MJ version 5.”
Alternatively, if you prefer to test the new model before fully committing to it, add the parameter “– v 5” at the end of your prompt.
Approaching Perfection
Midjourney unveiled version 5 of its commercial AI image-synthesis service, capable of generating photorealistic images at a quality level that has left some AI art enthusiasts uneasy due to its “too perfect” results. Midjourney v5 is now available as an alpha test for customers who subscribe to the Midjourney service, which can be accessed through Discord.
Graphic designer Julie Wieland, who frequently shares her Midjourney creations on Twitter, described her experience with the new version, saying,
MJ v5 currently feels to me like finally getting glasses after ignoring bad eyesight for a little bit too long. Suddenly you see everything in 4k, it feels weirdly overwhelming but also amazing.
Midjourney operates similarly to image synthesizers like Stable Diffusion and DALL-E, generating images based on text descriptions known as “prompts” using an AI model trained on millions of human-made art pieces. Recently, Midjourney found itself at the center of a copyright dispute involving a comic book created using earlier versions of the service.
Improvements in Midjourney V5
After experimenting with v5 all day, we’ve seen some incredible improvements, such as incredibly realistic skin textures and facial features; more natural or cinematic lighting; enhanced reflections, glares, and shadows; more expressive angles or perspectives of a scene, and eyes that are almost perfect and not wonky anymore.
Hands are Much Improved
Moreover, the new version has made significant progress in rendering hands, which has been a longstanding challenge for AI art generators. What we are seeing is that now hands usually have five fingers instead of the random assortment we saw in V4 and earlier.
Here is an example of a prompt that we entered into V5 and V4, again, you can see just how much more realistic V5 came out, and in particular, the hands look significantly better than they did in V4. In the second image (Midjourney V4) you can see her fingers look excessively elongated and rather strange.
Details are Enhanced
Another significant enhancement is the increased quality and refinement of background details and details that aren’t necessarily the subject of the image.
The AI image-synthesis service now generates more intricate and realistic backgrounds, providing a heightened sense of depth and atmosphere to the images. Below you can see the same prompt we typed in using V5 and V4. The first image was rendered in V5, and you can see that it is much more detailed. It also has a more interesting perspective.
While still cool, the second image is much simpler and less detailed.
This improvement in background details contributes to the overall realistic and immersive quality of the images produced by Midjourney v5, making them even more visually appealing and immersive.
It is Much Better at Creating Specific Characters
Another remarkable improvement in Midjourney v5 is its enhanced ability to create images of specific people with high accuracy. For instance, when generating an image of John Wick in v4, the result might have conveyed a general sense of the character that was reasonable, but upon closer inspection, the resemblance was not very accurate. Here is an example of V4. You can see his face doesn’t look quite right.
In contrast, Midjourney v5’s advancements have significantly improved the AI’s capacity to create images of specific individuals like John Wick, yielding results that are uncannily accurate and true to the original character. This heightened level of precision contributes to the overall realism and quality of the AI-generated images, making them more reliable and useful for various applications in art, design, and even marketing. Here is the same prompt in V5. You can see Keanu Reeves’ face is much more accurate. And, here too, you can see the improved background detail. None of our V4 Midjourney outputs for this prompt gave any background detail.
In the release notes shared on the service’s Discord, Midjourney highlighted that v5 now offers a “much wider stylistic range” than version 4. Additionally, it is more sensitive to prompting, generates less unwanted text, and provides a 2x increase in image resolution.
