Home Technical Talk

How much detail is too much?

I'm sort of a perfectionist, so I always run into this problem when modeling. How do you know how much detail to include in models, high poly models in particular?

For example, I want to do an AH-64 Apache for a project, starting with a high poly model. Is it realistic to try to include all of these details in the model? If not, how do you know what to include or not?

AH-64D_00-05207_11769.jpg

AH-64D_98-0123_6720.jpg

Replies

  • Rwolf
    Options
    Offline / Send Message
    Rwolf polycounter lvl 18
    Depends if your going for a accurate technical model yes. If your are doing a cinematic model and there is a close shot of part of that model, yes.

    If not, you should try to omit detail that could be done in texture maps. Look at the subject at a distance and judge what contributes to the silhouette of the model. Some obvious things to omit are rivits, bolts and nuts.
  • jocose
    Options
    Offline / Send Message
    jocose polycounter lvl 11
    I have to deal with situations like this a lot at my job. The work I do has to be technically accurate but its also for entertainment purposes so it can't take nearly as much time to make.

    The trick is to focus on the MAJOR design element. You, as the artist, need to pick that element out of the mess of shapes and then make sure that you polish that element and make it reasonably accurate.


    You don't compromise on your major elements and then make all the compromises you need on the other elements to get the project done on time.

    Then if there is a problem and they need you to explain why something isn't accurate you can simply or present you tell them that given items size on screen and the amount of time I was given it would have been a bad decision to spend the time figuring out how to accurately re-create that in 3D.

    Also try and get the BIG shapes into the engine as early as possible and then add details and make revisions after your initial export to the engine. You should always try to allocate time to make revisions and add additional details. If you made good decisions early on it should be easy to add details to your major elements.
  • Joshua Stubbles
    Options
    Offline / Send Message
    Joshua Stubbles polycounter lvl 19
    Depends on a lot of things.
    Game or render? Distance from camera? If it's a game, will the player ever get close to it??

    you don't want to waste time modeling parts of an asset if they're not going to be seen, or lost in the texture bake.
  • Sean VanGorder
    Options
    Offline / Send Message
    Its going to be one of the main pieces of a scene I'm going to do for class over the next 3 months. I'm going to start with the high poly and bake it down to bring it into Unreal.

    I'm required to have a video with it animated and moving, and I'm planning on having some of the shots fairly close up. I've been looking at assets from games I've played, and once I take the time to look at them up close, I realize that they are missing many of the details I see in real pictures. It's just that when you are playing you don't have the time to notice. I guess I'm just trying to find that line of when a player will begin to notice a lack of detail while playing.
  • leechdemon
    Options
    Offline / Send Message
    leechdemon polycounter lvl 11
    Personally, I've been using a fixed scale for general modeling. I've decided anything smaller than about 2-4 inches isn't worth modeling, but probably just textures instead. Some stuff that starts out seeming crucial will seem kind of silly when you use a subjective system like this. That said, most of our work is for driving simulators, and you're not out walking around inspecting stuff the way you would in an FPS.
  • pior
    Options
    Offline / Send Message
    pior grand marshal polycounter
    Yup you need to be very aware of your specs. Texture size, polygon count ... This is what will drive the level of detail of your object. It would be even better to have an idea the amount of pixels/feet or meters you are shooting for. Comparing with UT3 vehicles is a good idea!
  • nudity
    Options
    Offline / Send Message
    There is no such thing as too much detail, only time constraints. Maybe performance constraints too. if you have the time you might as well make it as high detailed as you please, even it some of the detail is going to be lost in UnrealEngine, you're going to use the model in your portfolio, no? just keep in mind your timelines and try to add detail around the whole model, rather than going through it one piece at a time, and try to focus on the high priority stuff (silhouette, body, etc) .
  • zahngol
    Options
    Offline / Send Message
    zahngol polycounter lvl 12
    This is kind of a re-hash of what jocose was saying but I feel that it bears repeating.

    Details are great, but make sure you get the major forms first. Start big, perfect the shape, then work inward to the next biggest shape. It's true that upon inspection, a lack of detail is noticeable, but if the whole form is mis-shaped and ill proportioned, that will stand out a whole lot faster.
  • Xoliul
    Options
    Offline / Send Message
    Xoliul polycounter lvl 14
    pior wrote: »
    Comparing with UT3 vehicles is a good idea!

    I'd say those are sort of the upper-limit in terms of modeled highpoly detail, Epic seem to not hold back at all there. But in tricount for the lowpoly meshes, they're starting to go a bit towards the low side.

    A lot of very good points here, but let me add that you need to consider the added time in every aspect of the creation process: those parts you're showing are not very "clumped together', more open and loose parts, with a very complex silhouette. If you go and model that in HP, you will need to create a fairly complex LP, and you'll need to put a lot more time in the baking process.
    So if you were to go full accuracy, you'll end up doing a lot of work that isn't necessary. Try to simplify in terms of silhouette, so that you can keep the number of seperate, small parts down in the LP.

    If you want to see how I did the high & low on a very complex object (applying what I just said here), check these out:
    Highpoly:
    http://cg.tutsplus.com/tutorials/autodesk-3ds-max/creating-a-next-gen-video-game-hot-rod-the-complete-workflow-day-3
    Lowpoly:
    http://cg.tutsplus.com/tutorials/autodesk-3ds-max/creating-a-next-gen-video-game-hot-rod-the-complete-workflow-day-5
  • metalliandy
    Options
    Offline / Send Message
    metalliandy interpolator
    Vassago wrote: »
    Depends on a lot of things.
    Game or render? Distance from camera? If it's a game, will the player ever get close to it??

    you don't want to waste time modeling parts of an asset if they're not going to be seen, or lost in the texture bake.

    +1
    I would consider planing out what sort of texture size the end product will be baked to as there is a finite amount of detail that can shown on any one texture.

    I always work on the rule that the finest amount of detail you can bake is 1 polygon per pixel, so assuming that you are using a 2k texture, the most detail you can accurately capture is 4194304 polygons (and that is when you are using 100% of the UV space).
    If we then say that you have 20% unused space, that gives you a 3355443 poly limit to work with :)
  • EarthQuake
    Options
    Offline / Send Message
    Rwolf wrote: »
    If not, you should try to omit detail that could be done in texture maps. Look at the subject at a distance and judge what contributes to the silhouette of the model. Some obvious things to omit are rivits, bolts and nuts.


    This sort of advice is given a lot, but i feel it is fairly misguided. Omitting things like nuts, rivets etc, things that are literally the absolute-easiest thing to create in 3d, and instead do them in 2d, where you have to worry about extra passes for masks, ao, and texture distortion. All you're doing is giving yourself MORE work by doing this work in 2d. Model one rivet mesh, use a "paint on surface" tool/script and fill your model out with rivets in a few minutes, get RBG, normals and AO data for free with no extra effort spent.

    So, to me looking at those sort of details, the ones that are very very easy to create and cutting those out is the wrong way to look at it. What you want to do is look more at:

    1. What areas are highly unlikely to be visible to the player, you can get away with simplifying these ares.
    2. What areas will have such fine detail if modeled accurate that they will not read well/have enough pixels to represent the forms, again simplifying these areas, this will not only speed up the modeling process but help to keep your work from turning into a mess of noise.
    3. Where can you cheat? Use floating geometry, create complex shapes using multiple intersecting meshes instead of large seamless meshes that will take forever to build. Separate your high into logical chunks, EI: real things are built from many small pieces, no need to keep everything one mesh.
    4. What areas are simply absurdly complex and not worth the extra time *even if you had it*?
    5. Which areas can you simplify, which shapes can you merge or simply change the design of to make the high work better with your low?

    Looking at it from this perspective it should begin to come clear how you can start shaving time off asset production.

    nudity wrote: »
    There is no such thing as too much detail, only time constraints.

    We can quickly get in trouble with this mentality. If you're adding detail simply to add detail and are unaware of the effect it has on style, scale, readability, cohesiveness in the scene/level/game, etc. Which are all very common flaws i see in a lot of work then yeah, you can EASILY have too much detail.

    MORE DETAIL =! BETTER
  • ralusek
    Options
    Offline / Send Message
    ralusek polycounter lvl 10
    i think our minds have been around game assets long enough that the brain just looks for consistency. after 5 minutes on super smash n64, my brain accepts the characters for what they're portraying, not the ridiculously low polycount. to maintain the illusion, just keep your details consistent. if you mix it up, it's just a reminder to the brain that something doesn't look right, even in terms of the game's reality.

    say you were just painting screws on a surface in 2d, and had a rivet that was even smaller actually modeled in right next to it, when the rivet catches light and the screw does not, your brain is going to be thinking something is fishy. if it was all just painted on, your brain has already accepted this consistent, albeit terrible, level of detail.

    cinematic, closeup, hero character, and objects directly interacting with the player character are the first place you'd probably want to make sure your detail is up to snuff. after that, keep the level of detail all relative to those, in order to keep things consistent.
Sign In or Register to comment.