Too much detail!

I often find myself feeling like the only way to make things look good is to provide a lot of detail in my models. For example, right now I am modelling rafters for an attic room, and the only way I can feel confident that they’ll look “right” is if I model the entire joint (I hate the messy geometry that booleans create) Granted, you won’t ever see the interior joint structure. It definitely feels a little silly, but, at the same time I’m getting a lot of visible detail in the model which would have been overlooked otherwise. As you can imagine, this philosophy of physically accuracy chews up a lot of time.

What is the best approach? I realize for gaming detail has to be “faked” a lot to manage consumer limitations. But for motion picture production, do they incorporate a lot of the same sort of fudging, or are all assets expected to have a fairly high level of accuracy?

Also, how do you personally balance your own tendency to obsess over details?

always depends if you need to do close shots or far away shots!

for close shots you might need to do high res models!

happy bl

If it’s just a hobby, it doesn’t matter. All you’re really impressing/pleasing is yourself. I mean, I’ve seen a lot of impressive 3d artworks to the point where I’m no longer in awe. Just saw the movie San Andreas and I can’t even detect where it’s looking fake. It’s as real as reality. And no one makes a big deal of it. Of course, I’ve seen lots of it in other movies prior to seeing that movie.

It’s a lot more practical now to model all of the detail because of the massive increases of RAM available for use in the render-farms used by studios.

It’s becoming more viable for consumers as well, today’s enthusiast-level hardware (the real high end stuff) allows you to pack up to 128 gigabytes of RAM inside of a single tower (and with Intel now coming out with 8-core processors), you can get the CPU power needed to render 128 gigs of detail to boot.

However, it’s still not going to be enough to render detail on the level that Hollywood does, Transformers 2 for instance has gained notoriety for having scenes that not only required over a terabyte of RAM, it created overheating issues in the best hardware available at the time (and that amount of power is not going to be in consumer-hardware for some time yet).

Hardware speed is not his main concern. His concern is the massive amounts of TIME it takes to construct details.

I would say it’s a factor though as far as time goes, the higher your hardware specs. the less in the way of shortcuts and optimization that you need to take (and it can take a bit of time to figure out just how to trim memory use and that doesn’t even mention the issue of Blender slowing to a crawl in areas like editmode and particle rendering).

Though Blender since the merging in of Bmesh has also gotten numerous enhancements in the modeling tools that save a lot of time as well (remember trying to model complex objects without the use of Ngons?)

Well, I do always try to keep my models as optimized as possible, and I intended this question as more general 3D than Blender-Specific. It should not matter if you’re in something as relatively inefficient as C4D or as massive as Clarisse or Arnold, keeping geometry density to a minimum is important for a number of reasons.

I’m not so much asking about density, but rather accuracy. When I model, I tend to think about how things are constructed, rather than how they look. For example, if I model an enclosure for a radio, I’ll think about how the case fits together. I end up with really accurate panels.

When I started doing 3d, I might have put down a box in the general shape of the radio, and added creases or extrusions that sort of represent the panels. I never really liked the look of it, topology tended to be pretty sloppy, and I did not feel things had a “material” presence about them.

Does that make sense?

I block it out first, and in some cases for non important props, they stay close to block when I’m sick of the project :). If I actually need the detail (such as mentioned radio case) as a main prop, especially in an exploded view, I tend to use CAD software. After sheet model complete and fasteners assigned, I may export to rendering for a quick remesh of the object.

I used to be “all the detail needs to be there, even invisible details and parts”, now I’m more “whatever gets the job done fastest way possible”.

I’d say it’s a question of what you get to see. If modelling the whole thing adds to the visual end scene, I’d probably do that too. If you don’t expect to ever see it in the finished render, I wouldn’t bother.

That’s only the case for things you won’t be reusing, though. If it’s a model you expect to use again and again, I’d go all out and get it as close to perfect as I could, as there’s no telling which angle I might want to use next time I use the model.

If I know that the details will not be seen in the final render then I don’t actually model them. Always cheat as much as possible is generally the motto in film production where everything needs to be produced as fast as possible.

^ Totally agree… :wink:

That’s why pre-viz & pre-production.
First make a visual block out, then decide for output resolution and apply knowledge you gain for final production (visibility, details, texture size, poly count…). For showing a detail, specific/separate scene is made. Optimize + optimize.
Gaming wise, use spectators POV (specs defined on platform to be developed for)… then create iterations/derivatives for different LOD.

@Shawn.Kearney - it’s really tempting to model things how they would work in the real world (guilty of it myself a lot of the time) but if it’s not useful to your current scene then avoid it for your own sanity.

With that said, maybe a realistically constructed model (even something as relatively mundane as a roof beam) would be useful to someone else. You could put these detailed models on blendswap or blendermarket or even turbosquid etc.

Thanks everyone. And yes. The roof beams was where I really started to realize things were getting out of hand. All that work and literally 10% of it is show, and none of what is being shown was all that hard work I put into it.