[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vr / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / asp / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / qst / sci / soc / sp / tg / toy / trv / tv / vp / wsg / wsr / x] [Settings] [Home]
Board
Settings Home
/3/ - 3DCG



Thread archived.
You cannot reply anymore.



I'd like to ask an extremely naive question about poly count. I'm new to 3D modelling, and I am just trying to understand the approaches to in-game character/asset modelling.

What's better: low poly modelling? low poly modelling with high res textures and normal maps? high poly modelling? High poly modelling with re-topology? Is there anything in between? Am I a total noob who needs to go do one hundred different character modelling tutorials before asking about "best" approach to poly count? Does it all depend on the "look" that I'm going for?

Thanks for indulging me in understanding anything that I can about the different modelling styles for game characters.

Pic related show three different "levels" of modelling, as I see it. Please feel free to elaborate on the approach taken to any of them. Thank you!
>>
>what's better?

It really depends what you're making, but most of the time for games, high poly models + high res textures = low fps. Unless you're going for a niche market, lower-poly models with quality textures and normal maps will be just fine.

Tutorials are definitely helpful. You don't have to even complete them, just make sure you're drilling the workflow into your head.
>>
>>602101

This was helpful. A little confirmation gives me some confidence that I understand some of what the heck is going on. thank you anon!
>>
>>602099
Over time the definition of what's considered "low" "medium" and "high" kinda changes as hardware does, so you have to consider things like the look you're going for and the target hardware. But generally speaking, low poly refers to an optimized mesh that can run in real-time on most hardware and high poly refers to models that aren't suitable for real-time use. Medium poly refers to a model that trades normal mapping for actual detail, while still staying optimized for real-time use. None of these refer to stylization at all, as you can have a very non-realistic game but still invest a decent sum of resources to define the graphics, or have something be highly realistic, but also minimal in geometric complexity.

It should be noted that when you consider things like poly budget, it's not just up to the target platform, but actually the level design. The level designer holds the keys to how much detail can be allowed in any given scene, as it's their job to do things like limit the number of things a player can see from any angle in space for culling purposes, spacing out convenient loading areas, and other optimizations that can affect how much you can have loaded in at once. It goes without saying that a game with a relatively locked-down camera can push a low more visual fidelity in any given view than an open-world game, where there is no realistic way to optimize for every possible action of the player.
>>
>>602105
very helpful, anon. Thank you very much.
Your explanation, along with explanations like these http://polycount.com/discussion/comment/2509602/#Comment_2509602 are giving me the proper insight and confidence in what I need to look into and learn.
>>
>>602105
It should be noted that modern game engines and computer hardware is very tolerant of transform workloads, that is raw geometry. If you have the option of not using a normal map by increasing an object's poly count by 30% or so, it can be a pretty decent trade-off depending on the resolution you were aiming for. If it meant saving on a 4k normal map, then by all means, have your environment piece have 12,000 tris instead of 8,000. There is never a point in this day and age where a barrel should ever have like 8 sides or something, be easygoing about it and slap 24 on.
>>
>>602110
roger that. Thanks again for the insight.
>>
>>602110
>>602105

I have a semi-related noob question: How much of my object (car, person, tree, etc) needs to be a single, continuous piece? For game use, does it matter if an object is ten different pieces put together? I guess this is the only way to make moving parts, like pedals?
>>
>>602099
>>602099
Don't overthink it. I just roll with whatever suits what I'm doing. For example, you'd worry about trying to keep as much detail as you can in as low a poly-count as you can if you were working on a game more than you would a film.
For the most part:
>Low poly
Has its own challenges. Try to put as much detail in as little as you can while having things process fast (prioritize render time, eg.) games). Can get a bit tricky to rig (turning the model into a puppet ready to animate) or UV-map (setting the model up to be textured) while minimizing distortion since there's not a lot of polygons to go around- you get less leeway in that regard.
>High poly
A tad more abstracted from the issues low poly deals with- resembles sculpting a bit more. You don't need to go crazy-high poly if you're willing to optimize. Some tricks here and there to have a lot of detail be a part of the material using some effects like normal maps and bump maps (textures that are used to make an object appear to have depth even if it's not a part of the model. eg.) cloth fibers).

And in the middle is the huge range between the two that can really lean either way. Don't stress it out too much, you'll learn the challenges each have as you go. If you're new, I'd recommend you learn by modelling both.
>>
>>602118
Doesn't matter unless you want it to bend (rigging and animation), or have it look perfectly seamless, or if it's just for organization.
It's not rare to have a bunch of separate models if you're making something mechanical.
>>
>>602118
Well, generally it's good to break an object down to a decent enough level for the game to be able to not render parts of it that aren't in view, but not too much as it increases draw calls for stuff that's not a static mesh. Where things really matter when it comes to multiples, are textures, as a single 4k texture is much more efficient than four 2k textures every time. Whenever possible try and fit the most amount of stuff into a single texture, some people even utilize the RGB layers of an image to store separate maps that the engine then splits up.
>>
>>602120
Absolutely. Thank you for lending your insight, Anon. Good stuff
>>
>>602131
>>602122

Thanks for the answers.

Do n-gons really matter if you aren't deforming something? Or triangles, for that matter?
>>
File: large gear.png (496 KB, 1332x945)
496 KB
496 KB PNG
>>602153
Absolutely not. You can have like a 20-sided gon for a wall or something and it won't matter. If it doesn't cause problems in-engine, than it's all good. In my own modeling style, I usually try and eliminate geometry that doesn't directly contribute to its shape. Pic related is the mesh for a large factory gear that I made
>>
>>602158
Not necessarily true. Reflections require good even poly spacing and consistent normal direction. If you have a mirrored surface Ngons will usually play havoc on it. But yea generally it doesn't matter.
>>
>>602170

Is this true even if you're working on something like a car? I don't know how you can model a complex, symmetrical object like that without mirroring.
>>
>>602179
you can, but you will be a moron for doying that....
how about modeling every tire separately?
>>
>>602153
>Asks if n-gons are a problem, as long as you're not deforming
>>602158
>Replies that n-gons can be okay, yet try and eliminate unnecessary geometry
>>602170
>Adds that n-gons can be a problem during mirroring, but are still generally okay
>>602179
>Follows up asking, how can n-gons be avoided when modeling a complex symmetrical object?
>>602184
Replies that you can use n-gons during symmetrical modeling. Implies moronic behavior. Confuses question at hand to be: efficiency of modelling symmetrically with inefficiency of modelling non-symmetrically. Derails actual question at hand: using n-gons versus avoiding n-gons.

OP here. Sorting that out, and clarifying for anyone else who got confused. [spoiler]... It took you 16 replies 4chan ... 16 replies :\ [/spoiler]

Thanks again to participants in insightful thread.
>>
>>602099
Depends. Personally I like to retopo a high rez mesh and then make an alpha map.

If you want a really detailed sculpt, like the picture you posted on the left: you need to do high rez, but you can still also retopo it later for rendering. The less polygons you have the faster your render time will be. Which is why:

If you’re trying to make video games: you need to know how to do low poly. Video games render in real time so you need to make sure nothing is slowing it down. Some AAA studios have powerful enough renderers to do high poly, like EA and their sports games. But in most cases you absolutely need to know how to utilize a limited amount of polygons.
>>
>>602099
Got a sauce on the middle image, anon?




Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.