how do you grade game art?
In a professional studio environment, What objective criteria would you use to grade a prop in a video game? Let's say it's a model of a chair, with a color texture map. If you had to assign a numbered grade to something, without using your opinion at all, how would you do it?
How do you decide that a piece of art for a game is not "good" or "fine" but "85%"?
This is a real question.
How do you decide that a piece of art for a game is not "good" or "fine" but "85%"?
This is a real question.
Replies
/thread
If I was a teacher I'd break it down, 20% for even use of topology and good edge flow, 30% for texturing, having no bad seams, and good use of UV space, etc etc.
Just my thoughts though.
does it look good?
I dont think in games, edge flow and UV space and all that REALLY matters in the final product. as long as it looks good enough to be in the game. obviously with proper edge flow and good uv layout and all that, it will have better chances of fitting the criteria, but from what ive heard, alot of shortcuts are taken as well to make things look like they should. may not be pretty technically, but visually, they are great.
At least not the looks of the art piece in question, obviously, if were talking technical stuff, like tri count, pixel density ratio, UVW space usage, scale, now that's a different case, and even that is pretty much down to logic:
- Will the player be able to see the asset very close to the camera?
- Is it a gameplay asset? (Needs clean and smooth collisions if player will walk on top of it)
- Does it reuse already existing textures / materials?
I could go on, and on.
Bottom line, if it looks good, it looks good. If it looks good and it runs great on consoles, then you'll make your coders happy! :P
I don't
It is based on your own observation as an artist, you decide if your art is good to pass on or not.
If you are working on an art assigned to you, you check to see if the requirements are met and if they are and they are looking good you pass it on to check if anything is missing.
@Yozora: for the first time I've seen you leave a topic by dropping 2 words...O.O
If there are any shortcuts to be taken, were usually talking about environment art. The shortcuts that are okay to take, should be of a visual nature, and never a wasteful on the performance end. I feel like its totally necessary to take shortcuts on the visuals because of the sheer volume of work that needs to be done for environment, if you treated every asset like a hero asset. The game will never get done. You need to re-use textures, maximize UV space, maximize your poly efficiency- which unfortunately doesn't look 100% awesome all the time. (but it can if done right)
the number one important thing in game art is efficiency. Everything else is secondary.
If you make an amazing looking asset, but its poly usage/distribution is crap, its unwrap is crap, and its pixel ratio is crap- at the end of the day, your beautiful asset is crap. Usually towards the end of projects there are pretty large optimization passes. and if a beautiful asset is asininely wasteful. it either gets the boot, or someone has to go in and 'fix" it which usually involves uglification.
However, I still dont think you can grade assets because context is key. how much is is used? how big is it? does it need to facilitate gameplay? etc etc etc.
At the end of the day, you can judge assets based on how intelligently they are put together.
Enough with the art is tough to judge and it is different for everybody mumbo jumbo.
That shit does not really apply here.
When creating pieces you go by more of a check list,
Does this wireframe and the amounts of tri's go? (what can i improve vs time)
Does this unwrap fit the needs and can i get away with it (what should i change vs time)
ect.. ect..
What more can i do with the texture and how much can i push it till i start spending hours that no one will notice (polish vs time)
Does what i create now, follow the lines of the concept and does this fit in with the other models.
All these things come with experience and asking.
Since everycase is specific, it is not just explainable with a few things you should do. Ask someone you work with, or your lead ect to have a look and give you a go or for tips and tricks.
Some pieces are created in a rush, some are done with a splendid polish,
It all comes down to weighing out as you go and that can only be done with exp.
1. Does it meet all the visual requirements set out for it ( style, level of detail etc)
2. Does it meet all the technical requirement set out for it ( tri count, texture size, optimization etc)
Try telling your lead that "art is subjective" when he/she asks you to change something......see where that gets you
This asset should have the following characteristics
polycount - no more than Foo tris, it should already be decimated
maximum texture size - 512k/1024k/2048k/etc square/rectangle texture in png or whatever format
channels - it should have diffuse, normal, spec, AO, etc
lighting - it should look good under full lighting, etc.
it should look good at close up or distant, etc.
it needs to look good from the following views, etc.
Possibly even have a test scene.
Rubric
does it meet the technical objectives - ie is it at or below the tricount, did the specified texture sizes get used
was floating geometry used where appropriate to reduce polys
has geometry duplicates been used where appropriate
was the texture space maximally utilized or are there large areas on unused texture space
were texture and polys allocated to where they would do the most good (ie were texture and polys wasted on non visible parts of the mesh)
how does the silhouette look for the views specified (are there unnaturally straight areas on an organic object)
are the textures crisp or blurry
are the saturation levels appropriate
is the texture 'too clean'
was tiling used - if so does it look natural or are their seams and obvious repeats
has lighting or shadows been 'baked in' if appropriate.
of course rigging, skinning, and other aspects need to be evaluated also.
There are lots more things to look at but that is off the top of my head.
I love you.
theres no inbetweens
Quality > Quantity :P
There is a techical side to what we do and thats what this thread is about.
hear hear
really though, if this question is being asked cos it's being put into practice somewhere (shudder) it makes me wonder what the sign-off policy is ...
"85% right, that'll do, box ticked, move on"
or
"only 15% more to go before you can move on ... 12% to go... 8% ... 1% ... done"
either one of those makes me feel a little queasy.
But you'll find it's pretty easy to just give everyone a very poor grade because by and large 90% of the students won't follow directions, won't spend more than a few hours at most on an asset that was recommended to have 10 hours plus on, and almost zero of the projects turned in will be aesthetically pleasing.
When your just in school it's just all about getting your stuff to look good and be on target. In the real world you have to do all that and deal with the needed of production and your team. I hate to say it but speed does matter, wiggle room for revisions does matter, and being able to understand different styles does matter. that kind of stuff really isn't covered too well in school from what i have seen. School is way more about the portfolio.
i.e. could someone else pick up the model file and your textures and be able to easily edit them? Did you name your layers/objects/PS layers? How destructive/non-destructive was your workflow?
Nothing worse than pulling up someone else's art to find everything named Box01, Box02 or Layer 1, Layer 2, Layer 2 copy, etc (or no source PSD file at all)...
The first day of (the ones I give) class, part of the intro material explains that each student must pick an art style and than all assets for the class must match that art style. They have to do a quick one week project putting together a bunch of found images in a similar style, be they photos or videogame screenshots, and then do some sketches of random found objects in that style. When I was grading I kept how well their assets matched their chosen art style in mind.
But I really cant stress enough how poor the quality of students will be. So far I've yet to hear of a single game art course that does any filtering past, "can you pay for this" and then flood through as many bodies as they can fit in the classroom. This results in 100:1 people who think the idea of working in games sounds fun to people who have an actual passion for making interactive art. With 20-30 people per class your chances of getting one of the good ones is low. Combined with the fact that even schools that aren't paid directly by the students benefit directly from the number of students in the school, and they will create policies that encourage very loose grading, meaning no bad students get flunked out or filtered. I can't count the number of atrocious students that through no amount of lifetimes spent in practice could ever contribute to a single videogame title in any way that would tell me I was the first professor to give them less than an A, and they were on their final year of a 4 year degree.
Lol yeah things like Box01, Box02, Spline78 can get annoying but it happens easy when doing large amounts of work quickly.
I love when at work and I open an environment to find everything creativly named yet have nothing to do with the obj at all. "Granny Beat Stick" "It puts the lotion on its skin" "lol naming is dumb" "chewing gum right now".
Always makes me smile a bit untill I need to find something.