How much detail?

Support Modeling
1

As written in the title. How much detail is enough? I tend to be a perfectionist, which doesnt mean my work is perfect. It’s just, that I am never happy with the results. It could be always better, more realistic, smaller, bigger, having more details or whatever else.

Part of my problem goes hand in hand, if talking about high or low res poly. I mean, a screw from far away could be made out from a few polygons and it would go as high poly. But if you look at the screw from a distance from lets say 5 cm, then i bet it would be considered as low poly.

But back to topic. How do you decide how much details a certain model should have? When do you use displacement maps or micro displacements to simulate details? I hope you have some answers which makes my life easier / happier :wink:

2

Answers? Do you often ask random people questions like “how much salt is enough in a meal” and then wink at them?

Because that’s what it reads, and the answer is the same: depends on what you’re making, along with the confusion about the wink.

Seriously though, there are many variables to it:

  • Design. If you’re modeling something that you don’t know the forms for, or don’t know what it should look like, it’s a design problem and you’re doing it wrong. Design stage is before modeling, and you should be either in that stage, or doing it by gathering references
  • The target use: still, animation, 3d printing, game engine, other. The requirements change between those, and also within each. Hero game asset is often different from a background prop, the performance might not be good enough for the detail when X amount of assets needs to be drawn. 3D printing can only use geometry to define forms. Animation may need rendering optimizations, deformations might need to be handled in a certain way
  • How the model or models are used in the pipeline http://www.upcomingvfxmovies.com/wp-content/uploads/2014/03/3d_production_timelines.jpeg before getting to the end result
  • Distance to the camera. Geometry and textures are optimized for viewing them from a distance. The more distance increases, the need for detail decreases. Movies use matte paintings for backgrounds, set extensions, and set detail in compositing. Games use LOD levels and mip maps, and also fake geometric detail with textures. Human eyes and visual processing drop detail
  • Silhouette. Forms against the background or against other forms on the model itself need to be defined with real geometry as bump/normal mapping doesn’t work for that. How much density the geometry should have depends on the curvature of the form and the view distance.
  • Style and other assets. Foreground, middle ground, and background assets should have decreasing level of complexity to look realistic. The assets on same level should have similar refinement on their details. Style choices might make an exception or disregard that altogether
  • Render resolution. No need to put geometry and/or texture detail if there are not enough pixels to capture them.

There’s more but that ought to be enough for this answer.

3

Thank you for the fast respond. It’s very detailed. Most helpful to me was the fact about bump /normal maps against background.

But I think, your answer is more for the big pictures, while im searching an answer for the small pictures. Let’s say, I do want to create just a table, a chair, or a street sign. Just a simple object which I’ll put in my folder and use them as assets.

Talking about chair or table. I could use those two objects within a scene in a dinning room where I dont necessary need a closup of them. But using them as a construction manual I might want to see much more details like screws in table and chairs and such.

Well, maybe I’m completely wrong in my thinking and creating such assets is a waste of time? Should I rather create things from scratch, and depending on the scene and target I choose the details for each object?

I feel a bit lost, since I always try to salten up my meal more and more and dont know when to stop.

4

If the question is how much detail you put in an asset so it can be reused, then all the details in my previous answer apply. You can’t prepare a model that is suitable for everything. The requirements for different purposes are different for visual reasons, and because of the pipeline.

What you can do is keep consistent detail level, model structure in such shape that it’s easy to modify, real world scale for finished assets, good file and asset organization within the files, and object origin in such place it can be positioned easily after adding it. For a table, chair, or even humans or animals that would mean the origin is in between the legs and on the ground.

For a scene that has tables and chairs you might need to change the size, color, materials, material assigned textures, and that’s assuming it’s the type and style of tables and chairs the scene needs. The only part visible of screws probably are the screw heads, which might be a texture detail because they have such unimportant but still visible part in the scene.

A construction manual has different visual requirements. Have to be able to visualize the placement of parts and their relationships quickly, which could mean simplifying the render to something non-photorealistic, like outlines with barely visible solid faces.

If you ever been to the hardware store, you know the collection of screws, nuts and bolts, different heads is ridiculous. The construction manual only has to visualize them enough to be able to differentiate them from other fasteners. That doesn’t mean you actually model a screw with actual length, actual thread pitch, actual head, just enough to know which is which within the build. Text and numbers tell which it actually is and what size tool to use.

Good designs use as few different fasterners as possible because it makes economic sense, and is easier to construct. Extra objects addon includes screws and nuts. Could be more comprehensive but still helps a lot. Helped me when I illustrated build instructions.

5

I completely understand of what you are talking. Maybe I’ll stick with a consistent detail level.

Assumed we take the chair, which I created as an asset and used it in a certain scene. Now I want to create another scene, where you can see an ant crawling over a screw of that chair. In that case I would take my chair asset and make just that screw part of the chair more detailed?

Overall said, the hard part for me is to decide when to stop with the details, since I mostly think at creating assets for whatever needs. That’s because I tend to tell myself that I dont have enough time for a whole scene, so I create smaller objects which I eventually can use in a scene. But I guess, that seems to be the wrong approach then?

6

Exactly.

No, just a very specific scene and needs. And even if you were to only do realistic interior renders, could prepare it for one scene and maybe use as a side prop in other scenes, but you’d still want variety of options. That’s why sites that sell assets have all kinds of packs available, which might need tweaking to suit your needs. Can’t build a traffic that only has one type of car. Well, unless it’s a Porsche motor city, could use one model and just change the color.

7

I see. Makes all sense. I just have to work on myself then to figure out, what exactly I want to do.

Thank you very much for your patient answers.