Discover Midjourney Mastery! Immerse yourself in the comprehensive Midjourney course on Skillshare. Gain a deep understanding of the Midjourney process, refine your skills, and achieve remarkable results: Join my class on Skillshare and get 1 month of Skillshare free.
I vividly remember my first interaction with Midjourney in September 2022. Unfortunately, I can’t recall my exact first prompt because I initially used my friend’s account, but I can show you the first prompt I entered on my personal account I created on the same day.
The prompt was:
“Lao Tzu eating Lasagne, hyperrealistic, shining color.”
And this was the result after doing a few variations:
From today’s perspective, this may not seem too impressive, but at that time, I found it revolutionary. This image sparked a sense of wonder within me. It offered a glimpse into the future, it was a primitive vision of what Midjourney could later become.
Today, only 8 months later, we already take so many things for granted that were mind-blowing just a couple of months ago.
That’s why in this post, I want to explore the evolution of Midjourney since its beginning in March 2022. I will show you a couple of prompts tested with Midjourney versions 1 to 5.1. Enjoy the ride!
These are the 3 prompts I am using for this test:
- A sad astronaut, walking lost in space
- Lao Tzu eating Lasagne, hyperrealistic, shining color
- Female and male, couple, striking eyes, soft lighting
By the way, you can test the different versions of Midjourney yourself by adding the parameter "--v" (version number) at the end of your prompts. For example, "--v 1" would create an image with the first Midjourney version.
A sad astronaut, walking lost in space –v 1
It looks as if Midjourney v1 understood this prompt. Though, the result is still a bit random.
A sad astronaut, walking lost in space –v 2
Midjourney v2 Looks already much better. The astronaut now stands on 2 feet, but what is this balloon doing in the background?
A sad astronaut, walking lost in space –v 3
In Midjourney v3 we even got our astronaut with a face that clearly reflects a sad impression.
A sad astronaut, walking lost in space –v 4
Midjourney v4 made a quantum leap forward. Now, the image was starting to look photorealistic,
A sad astronaut, walking lost in space –v 5
Midjourney v5 interpreted the prompt as a highly detailed comic drawing. As in v4, the image is lacking the sadness.
A sad astronaut, walking lost in space –v 5.1
This is v5.1, the latest and most advanced version available. The image is highly detailed and evokes a sense of cosmic horror (but this might be my interpretation).
Lao Tzu eating Lasagne, hyperrealistic, shining color —v 1
Midjourney v1 managed to generate a lasagne, however, old Lao Tzu is missing entirely.
Lao Tzu eating Lasagne, hyperrealistic, shining color —v 2
In the second version we finally got Lao Tzu, however, this looks more like some gross nightmare than the Chinese philosopher eating lasagne.
Lao Tzu eating Lasagne, hyperrealistic, shining color —v 3
This is the version with which I was starting. Here, Midjourney begins to put the different parts together. However, the results still look a bit random.
Lao Tzu eating Lasagne, hyperrealistic, shining color —v 4
Again, we see that the update to Midjourney v4 was a huge improvement. Now we got a highly detailed Lao Tzu. The only question remaining now: Is this really lasagne he is eating?
Lao Tzu eating Lasagne, hyperrealistic, shining color —v 5
Midjourney v5 improved the results even more, but the lasagne was exchanged with pasta in 2 of the 4 images.
Lao Tzu eating Lasagne, hyperrealistic, shining color —v 5.1
Interestingly, as Midjourney progressed to its latest version v5.1, the depiction of Lao Tzu became more refined and detailed, but it is still not getting the lasagne right.
Female and male, couple, striking eyes, soft lighting –ar 2:3 –v 1
This is an attempt to create realistic photography of a female and male couple in Midjourney. V1 fails at this attempt, however, it could pass as some sick modern art.
Female and male, couple, striking eyes, soft lighting –ar 2:3 –v 2
Midjourney v2 started creating NSFW pictures. It still looks totally random and strange, but it is nevertheless a huge improvement compared to Midjourney v1.
Female and male, couple, striking eyes, soft lighting –ar 2:3 –v 3
V3 even generated faces. But they look alien and give the viewer a chilly uncanny valley effect.
Female and male, couple, striking eyes, soft lighting –ar 2:3 –v 4
This example proves that the progress from Midjourney v3 to v4 was a remarkable advancement in Midjourney’s capabilities.
Before, Midjourney merely generated some strange AI collages. Starting with v4, we get images of people that could easily pass as computer game characters.
Female and male, couple, striking eyes, soft lighting –ar 2:3 –v 5
Midjourney v5 added even more realism…
Female and male, couple, striking eyes, soft lighting –ar 2:3 –v 5.1
And here is the cumulation of the efforts so far. Midjourney 5.1 produces results that are actually hard to distinguish from real photography.
Conclusion:
Reflecting on my first interaction with Midjourney back in September 2022, I am reminded of the revolutionary impact it had on me at the time. The initial image generated, though primitive by today’s standards, ignited a sense of wonder and sparked my curiosity.
In just eight short months, Midjourney has undergone remarkable growth and evolution. From its early versions, marked by randomness and novelty, to the latest release of v5.1, the progress has been astounding. The improvements in detail, realism, and the ability to understand complex prompts have been nothing short of a quantum leap forward.
My Midjourney-related projects
- Discover Midjourney Mastery! Immerse yourself in the comprehensive Midjourney course on Skillshare. Gain a deep understanding of the Midjourney process, refine your skills, and achieve remarkable results: Join my class on Skillshare and get 1 month of Skillshare free.
- Create better prompts with my Midjourney Prompts Generator for Notion
- Check out my best prompts on Promtbase.