[Advanced Tutorial] Midjourney 2D-3D evolution process
Keywords used: unreal engine virtual engine
Game rendering engine, the latest version is Unreal Engine 5, which has 3D effects to increase the fidelity of graphics. There doesn't seem to be much difference between adding the number 5 and not adding the effect.
It is often used with "Octane Render" and "Lumion Render", "Octane Render" is a modeling rendering, "Lumion Render" is an architectural rendering, both can increase the realism of the image. "The former is used for modeling rendering, and the latter is used for architectural rendering."
So, let's get started.
The simplest idea: a black monster with a flaming tail.
Black monster, tail on fire, --ar 2:3
Feedback after handing over to MJ. Black, monsters and fire, no problem with AI. Then generate and view again.
The monster got bigger and the background became darker. There were three light-colored backgrounds and one dark-colored background before. It is estimated that AI thinks dark places are more scary. But it’s a bit incomprehensible, and we still can’t be lazy. It’s no initiative to leave it to MJ completely. What we have to do is to control the AI and let it go with us.
Black monster+big eyes+sharp teech,tail on fire, --ar 2:3
Of course, simply add "big eyes" and "fangs" and look down.
Sure enough, the eyes are big, cute, and fierce. Are you here to show off the cuteness of V2? I'm drawn to this, but it's okay to be cute. Who said big black monsters can't be cute?
See what else the V2 can become.
MJ seemed to feel the compliment, and stared directly at it with four pairs of big cute eyes, and chose the fluffy fluttering shape, which echoes very well against the background of the fire, so he decisively chose U2.
The next thing to do is add "Unreal Engine" and see how this keyword turns 2D into 3D.
(Of course, we could have added it in the first place, but then we wouldn't have been able to see the evolution, and wouldn't have known exactly what the word "Unreal Engine" was doing)
unreal engine, black monster+big eyes+sharp teech, tail on fire, --ar 2:3 --seed 2827
Now with Unreal Engine added front and back, it's fine in many cases. But if you have a lot of descriptions, it's better to add them later. Because AI recognizes the description, the front part is more important. Of course, there are slight differences between V3 and test/experimental modes.
OK, here are our results.
Wow, 2D has really become 3D, even though it's just an Unreal Engine, the change is still very noticeable. The original detail has increased a lot, and the depth of field is more three-dimensional.
Next is U2.
The head is already fine, the body will be done next time, because I didn't add more descriptions.
The overall effect is good, the original idea has basically been realized, and MJ's algorithm and game performance are commendable. Most importantly we know what the word Unreal Engine can do.
At the beginning of the game, I basically just copy the description of a good picture and generate it myself to see, and nine times out of ten it turns out to be bad. One possibility is that there is a lot of back and forth in the process or you have to repeat patiently; the second is simply copying a bunch of words without knowing what each word means or what effect it has, so it is very passive.
So, with the help of a simple idea, try a certain keyword and have a pretty good idea of what you want.