#which you don't really notice until you SEE a designed game. like cp2077

[zevranunderstander]

okay im off work now, so i can explain: basically for most jobs in the games/film/animation industry the art isnt just there to give you the *aesthetic* of a thing. I mean, yeah, you might also want to convey a *vibe* but that is the LEAST of what you want to do. any given moodboard made by a completely unskilled person can convey a vibe. concept art has a LOT of thoughts behind it - if you just wanted to do storyboarding you didn't need a fully rendered picture in the first place because that's just a massive waste of resources.

so what DOES a concept artist (let's say for a game) think about when making a concept? This is probably not everything but just from the top of my head:

What is the scale of everything? How does the player character scale compared to each object? What are the exact measurements of objects? (you need to model all of them, you need pretty precise details when it comes down to it) What ramifications does the the game camera have on the scale of the level? (= If you want a really tight space, the camera would constantly clip through walls, so you need to adapt space to common game design knowledge. Other examples of this would be that doors usually need to be a lot bigger than they are irl without looking weird or that if your character moves reeally fast through the levels, small details become completely confusing and unrecognizable, so they need to be WAY bigger than they are irl without looking goofy)

What kind of design/style choices need to be applied to this image? Does the game have a specific color scheme? How does worldbuilding influence the environment (building styles/specific types of plants/fonts used in the world/etc)?

What kind of technical ramifications do you need to consider? if you create an environment you need to think about having an asset set that is reusable (= how often can the same tree model be in this scene without people noticing)? does modeling each of these objects work? how to texture these efficiently (texture is like, a whole post in itself)? can the team's shader artists/the engine actually DO this effect?

and MOST importantly: what are your thoughts on level guidance in a shot? Is it readable where the player can step and where they can't? Is it readable where a player would die if they would step there? Is it readable where the player is supposed to go? where is this shot you just made on the map? what other landmarks can the player see from this place?

i want to do this on an example, so take this piece of concept art for fallout new vegas:

this probably took an artist 1-2 days. an AI would be way quicker, right? So what does it matter that the artist put more thought into this, you might ask. sadly i don't have access to midjourney right now to try and recreate this in there, but i tried to find something at least remotely similar on google right now (and i tried to be fair here, and tried to find something that is actually good and not the super unspecific images you get from midjourney *most* of the time when you ask for something complex):

like, the first image is basically production ready, you give this to the team and the 3D artists can think about what models to build out of this so they can reuse as much as possible, the picture shows what textures are needed, its shows how the town should be built up, etc. etc. etc.

it is also worth to note that the image shows Primm, one of the locations in fallout new vegas, which is based on a real small town in Nevada. you can see the visual research that went into this, as this is clearly a post-apocalyptic version of real-life Primm, not 1:1, but adapted to the game's level design. It has the sights Primm has, it shows the quest-relevant locations, it shows WHERE Primm is located and what the surroundings look like - the desert is intentionally flat, because the designers of the game wanted the Lucky 38 tower (seen in the background) visible from almost every place in the open world, because of how central it is to giving you a sense of location and a goal to reach.

And I cannot talk about EVERYTHING on here but this picture holds a lot of information. It shows us exactly what material everything is made out of, it shows us where a player can reasonably stand and where not, it shows us how big this settlement is supposed to be. I will quickly concentrate on the "Motel" sign in the foreground, to show what this image does in a microcosm:

the sign shows a type of sign design that fits into new vegas's style identity of having a 40s vibe

it indicates that this place used to have a motel but the sign wasn't needed anymore

there having been a motel means that this place was probably close to an open road/highway (it is)

the way it is placed indicates that it is used to reinforce an old fence, meaning that the fence in this picture is not just an ancient remnant of Primm, but something that still protects people behind it

...and so on. and you can probably find that much thought in a lot of these props.

now look at the midjourney one. the ai doesn't have any context for the world. you would need to give it *all* that information - for every attempt to generate an image - the story of your world, the concrete setting, the 40s influence, the specificities of the town Primm, the overall world design and conceptualization, the idea of texture and asset modularity, every small detail about the motel sign would be needed to be described. it would need to build a logical structure for a town, and so on and so forth. even if you wrote a perfect description of this picture, midjourney wouldnt be able to be that precise in what you really want to see. it's good for getting a pretty picture, but it's not very good at giving you a specific picture you want.

i will not go into this too long, but look at the midjourney image and imagine you have to make a video game scene of this image, the more you look at it the less sense it makes because midjourney doesn't understand what it is showing per se. where should the player walk and where not? what are ANY of these textures supposed to be? what can the signs tell us about this place? *where* is this? what even *are* most of these items? - answering all of these questions is EXACTLY what an artists job *is* in a game. generating this brings you about 5% further in conceptualizing the world, because you still don't have designs for signs or props or anything. you don't need a *vibe* when making a video game

and managers and CEOs who know jack shit see this now and go "oh? a way to save money and be more efficient?", while having NO idea that the program is ENTIRELY unsuited for the job. like, midjourney couldn't even produce a consistent style for a visual novel where you would get the same character with different expressions in the same artstyle, it can't produce a consistent style in UI elements, i can't do concept art, it can't do any asset sheets that are more unique than a set of rocks. it might be able to give some general inspiration and it might be able to be used as a very rough base for painting over to save some time & already have an atmosphere (artists already do that, thats why its also called "photobashing"), but outside of that it is practically useless

as someone in the games industry and as someone who tried midjourney id say i give it about 5 months until project leaders come crawling back to beg to rehire all the artists they fired