Are you an artist? Do you like to create digital art? If you do, then you’re going to want to check out the new Midjourney V4 base model.
With this new V4 model, you can create much more realistic imagery than ever before. Plus, the Midjourney team has been working hard to improve the resolution, quality, and controllability of upscaling.
So if you’re looking for a way to take your digital art to the next level with Generative AI, then you definitely need to check out the new Midjourney V4 base model.
Read more, or watch the YouTube video (recommended):
YouTube:
What is Midjourney V4?
Midjoruney V4 is an entirely new codebase and totally new AI architecture. It’s Midjourneys first model trained on a new Midjourney AI supercluster and has been in the works for over 9 months.
V4 is not the final step of Midjourney`s AI Art generator, but what they consider a first step.
Meaning that the Midjourney team hopes this V4 version will feel as the new beginning of something deep and unfathomable that can change how humans create design.
The V4 model can generate much more realistic imagery than anything Midjoruney has released before. This can also be great news for everyone trying make a business with AI art.
From now on they will be much more strict with our rules around the “don’t be a jerk or create images to cause drama” policy.
What’s new with the Midjourney V4 base model?
The V4 model from Midjourney is aiming to improve in a lot of areas when it comes to producing more precise and fulfilling the Midjourney prompt author’s imagination.
Some of the new things we know so far are listed below, and I have tried to put all of these categories to the test with my own prompts:
Vastly more knowledge (of creatures, places, and more)
To test the vastly more knowledge of creatures, places, and etc i did craft the following two prompts:
a huge blue dragon, spitting fire on the eiffel tower, fantasy, digital art, sharp focus, –v 4
a large werewolf roaming the streets of Manhattan New York, people running away, digital art, –v 4
My thoughts here were that the Dragon and Werewolf was the creature and Eiffel Tower and Manhattan were places.
This is the results i got back from the V4 model which i was pretty happy with and clearly shows some improvement:
Much better at getting small details right (in all situations)
To test the category of getting small details better i did craft the following prompt:
teenage boy watching the show “teenage mutant ninja turtles” on his TV in his basement, 1980`s, digital art, sharp focus, –v 4
Here I wanted to see the detail “teenage mutant ninja turtles” on his TV. The results I got back here was kinda exactly how i pictured it, so very happy with this prompt:
Handles more complex prompting (with multiple levels of detail)
In the complex prompting category I went over to OpenAI`s GPT-3 to ask for a complex prompt describing a Cyborg in extreme detail.
This is the long and heavy descriptive prompt i got back from GPT-3:
The cyborg is a woman who has been outfitted with a series of robotic implants and artificial limbs. Her body has been extensively modified with cybernetic devices that allow her to perform feats of strength and agility that would be impossible for a normal human. She is equipped with a wide range of weapons and devices, and her body is protected by a suit of armor that is nearly impenetrable. She is a fearsome opponent in battle, and she is able to think and act at lightning-fast speeds, –v 4
I gotta say i am really happy with the result i got back from Midjoruney V4 here that you can judge for yourself here:
Better with multi-object / multi-character scenes
In the multi-object / multi-character scenes category I again went over to OpenAI`s GPT-3 to ask for help. I asked for:
Describe in EXTREME DETAIL a multi character scene from The Viking Age, set in Norway in a realistic digital art style:
The prompt i got back from GPT-3 was:
The scene is set on the coast of Norway. There are two characters: a viking warrior and a dragon. The viking warrior is clad in traditional viking garb: a fur tunic, leather pants, and a metal helmet. He is armed with a sword and axe. The dragon is a massive reptilian creature with leathery wings and razor sharp claws. The two characters are locked in combat. The viking warrior is swinging his sword wildly, while the dragon is using its claws and teeth to try and attack the viking. The ocean is churning behind them and the sky is dark and stormy, –v 4
Also here i got to say the new V4 model really pulls this off, just look at the result from this complex prompt:
Supports advanced functionality like image prompting and multi-prompts
The last thing I wanted to test from the V4 model was multi-prompts. In this case I used the prompt:
A two-headed coin is an extremely rare object because it is a physical representation of chance and good luck. A four-leaf clover is an extremely rare object because it is a physical representation of luck and good fortune, –v 4
This multi-prompt feature takes two different prompts and merges them into one as far as I can see. In this example you can clearly see the two-headed coin and the four-leaf clover merged together into one coin.
So I was very happy how this turned out!
Remix mode on Midjourney V4
The Midjourney team is also rolling out remix as fast as they can on this new base model.
– To enable type /settings and click ‘remix mode’
– Now when you click variations you can change the prompt (or the mode from –v 3 to –v 4 etc)
What’s new with the V4 upsampler?
The new upsampler in the Midjourney V4 base model adds, changes and corrects details in the original image (more like the v3 upscaler).
Also the new upsampler is an early work-in-progress and will change significantly over time.
What are the Midjourney team still working on?
The Midjoruney team is always working on improving their AI model, some of the things the team is working hard on now is:
- Increasing the resolution, quality and controllability of upscaling
- Adding more custom aspect ratios (currently limited to 1:1 in V4)
- Continue to increase image sharpness and reduce instances of tight cropping
- Greatly reducing text artifacts (certain prompts become unreadable language on images)
- Improving the default styling
Conclusion
The Midjourney V4 base model is a huge step forward for digital art. With its new and improved features, you can create much more realistic and detailed images than ever before.
Plus, the Midjourney team is constantly working to improve the resolution, quality, and controllability of upscaling, so you can be sure that your digital artwork will only get better and better.
So I think you should really check this out now.
